Identifying and analysing carbon hot-spots in an Inter- Regional Input Output framework

Transcription

1 Identifying and analysing carbon hot-spots in an Inter- Regional Input Output framework Antonios Katris Ph.D. in Economics and Environmental Science Division of Economics 2nd November 2015

2 Thesis Abstract Input Output frameworks have been widely used to study the emissions of industrial sectors either in specific economies or globally but usually focus on aggregated measures under production or consumption accounting principles (PAP and CAP). This leads to a lack of transparency in terms of the structure of the emissions and provides limited information on what are the main drivers of the emissions allocated to each sector under PAP and CAP. This information gap limits the options of policy makers to interventions on whole sectors, rather than the components of their supply chains that hold the major shares of the total embodied emissions. In this thesis we argue in favour of a more disaggregated, a hot-spot, approach that provides a better understanding of the structure of emissions under both of these headline measures. We develop a methodology to identify CO 2 hot-spots in downstream and upstream supply chains, both domestic and global. The methodology is applied first to a Single Region Input Output framework for China in 2005 identifying Electricity, Gas and Water Supply as the Chinese sector with the highest direct emissions. Examination of the sector s domestic downstream supply chain reveals that the majority of emissions are generated to support the final demand of other domestic sectors. Of these Construction is the main driver and it also is the Chinese sector that is found to have the largest domestic CO 2 footprint, with several emissions hot-spots in its domestic upstream supply chain. The hot-spot methodology is then extended to a global Inter-Regional Input Output framework to consider hot-spots in a global supply chain context. By focusing on 2009 (the year for which appropriate data are most recently available) and UK total final demand we find that Chinese Electricity, Gas and Water Supply is the non-uk sector with the largest direct emissions driven by UK total final demand. Studying this sector s downstream supply chain outside China reveals that a large share of the sector s emissions is ultimately generated to support several UK-based sectors domestic final demand. Furthermore, the UK Health and Social Work sector is identified as the UK 2

3 sector with the second largest global footprint to support domestic final demand. We identified a number of hot-spots in the international part of its upstream supply chain, with a key finding being its dependence on the activity and the embodied emissions in global chemicals production. Finally, the thesis goes on to demonstrate how conducting hot-spot analysis on disaggregated regional/subnational Input Output tables can provide more detailed local level analysis of hot-spot findings from the Inter-Regional Input Output framework. The key finding in this respect is the importance of introducing region specific emissions data where possible, as non-region specific data can lead in incorrect estimation of the embodied emissions in any component of the supply chains of any sectors. In general, through this research project we developed a methodology that can enhance the policy makers understanding of the structure and the drivers of the emissions generated throughout the economy. This additional information on the emissions structure, when combined with familiar IO analysis on employment and value-added for example, has the potential to lead to more targeted/focused policies, which result in significant emissions reduction with the minimum employment, resources and valueadded cost. 3

4 Table of Contents Thesis Abstract... 2 Chapter 1: Thesis Introduction Policy background and relevance of thesis Structure of thesis Chapter 2: Identifying CO2 hot-spots in Input Output frameworks: A Chinese single region example Abstract Introduction Hot-spots and key sectors Methodology Single Region Input Output Environmental Input Output Hot-spot detection Data Results of hot-spot analysis Direct emitters Emissions attributed to final demand Conclusions and extensions Chapter 3: Carbon hotspots in global supply chains: an inter-regional input-output analysis Abstract Introduction Literature background Methodology and data Inter-Regional Input Output Environmental IRIO Emissions hot-spots Data Results General overview Type (a) and downstream Type (c) hot-spots outside the UK driven by UK total final demand Type (b) and upstream Type (c) hot-spots driven by UK total final demand Conclusion and extension Chapter 4: The benefits of using regional Input Output tables and the importance of region-specific satellite emissions data Abstract Introduction Literature background Methodology Environmental Input Output Emissions hot-spots Data Case study demonstrating the benefits of using SRIO based on national/regional economic IO data to add information to an IRIO hot-spot analysis Added value from introducing region-specific data for Scotland Emissions reported by Scottish industries The importance of region specific satellite emissions data Conclusions and extensions

5 Chapter 5: Thesis Conclusion Acknowledgements References Appendices Appendix 1.A: The Industry Industry industrial sectors of WIOD database Appendix 2.A; Table 2.A.1: The countries included in the OECD Inter-country Input Output Database Appendix 2.B; Table 2.B.1: The Industry Industry industrial sectors of the OECD Inter- Country Input Output Database Appendix 2.C: The creation of satellite emissions account C.1 Allocation of emissions from fuel combustion C.2 Allocation of emissions associated with autoproducers C.3 Fugitive gases and industrial processes Appendix 3.A: The countries included in the OECD Inter-country Input Output Database Appendix 3.B: Scottish sectors matched to OECD sectors Appendix 3.C: The CO 2 emissions intensities of Scottish sectors Appendix 4.A: List of abbreviations

6 Chapter 1: Thesis Introduction 1.1 Policy background and relevance of thesis Input Output (IO) analysis is a widely used tool in the research of the CO 2 emissions generated due to economic activities. Especially over the last two decades, a large number of studies have been published that calculate and study the total direct emissions 1 of each industrial sector in different countries, including among others Brazil, China, Denmark and USA. The total direct emissions reflect the emissions generated by each producing sector to support the final demand of all the other sectors, not just their own final demand. This being the case it was recognised that the final demand of all the sectors in an economy has a significant role to play in the emissions generated directly by any sector. As such researchers used IO analysis to also calculate the CO 2 footprint 2 of all the sectors in different economies. In fact, most studies calculated both the total direct emissions and the CO 2 footprint, comparing the two (see Wiedmann et al, 2007 for review). This thesis sets to address a problem that exists when focusing on the calculation of these headline figures. As can be seen by examining an IO table, each producing sector sells different value worth of output to each consuming sector and in much the same way each consuming sector requires different value of output from the different producing sectors to meet its final demand. Under the assumption that each producing sector charges a flat price of output to all consuming sectors, the differences in value of output paid actually reflect differences in volume of output required. Therefore, it can be seen that each producing sector distributes different volumes of output to each consuming sector and vice versa. The methodology proposed in this thesis disaggregates the downstream and upstream supply chains of each sector, and therefore the headline figures of total 1 Direct emissions are the emissions generated by any sector to meet its own final demand and support the final demand of all the other sectors in an economy. 2 In this thesis the CO 2 footprint is defined as total volume of CO 2 emissions embodied in the output of all the sectors featured in the upstream supply chain of any given sector, acting as direct requirements to meet that sector s final demand. 6

7 direct emissions and CO 2 footprint, allowing the study of the structure of the supply chains and the embodied emissions in each component of the supply chains. The development and results of IO analysis tools are closely related to policy. For instance, the recent international climate change mitigation agreements under UNFCCC (Kyoto 1997, Copenhagen 2008, Paris 2015) or Europe 2020 strategy, employ a goal setting approach where each participating party is required to reduce anthropogenic greenhouse gas (GHG) emissions below the emissions of 1990, which is set as a baseline year. Each party is held responsible for the emissions generated within its territory under a territorial Production Accounting Principle (PAP), i.e. the total direct emissions of the producing sectors in this party/country. Researchers have argued, even before the implementation of the Kyoto protocol in 1997 (see Wyckoff and Roop, 1994), that such an approach would limit the effectiveness of any international agreement as parties would be in a position to meet their emissions goals by shifting the most polluting parts of their production to developing nations without strict environmental regulations and/or emissions goals, i.e. pollution leakage (Arrow et al, 1995). An alternative approach that has been developed (see for example Munksgaard and Pedersen, 2001), involves a Consumption Accounting Principle (CAP), under which the consumer is held responsible for the emissions generated to meet its demand. The CAP emissions of any given sector include the CO 2 footprint of that sector plus the emissions generated by consuming the final goods produced by the sector under examination. Consumption oriented approaches have also been investigated by governments, including the UK, as potential ways of achieving their emissions goals. A report by the House of Commons Energy and Climate Change Committee in 2012, reflects on the potential benefits that the committee sees in adopting consumption-based approaches. However, one of the main issues that have been identified involves the lack of credible inter-regional data, which would be crucial in studying CO 2 emissions under CAP as quite often parts of the upstream supply chain of any sector are located outside the territorial limits of each country. This problem has been flagged up by the House of Commons Energy and Climate Change Committee (2012a) but the development of the World Input Output Database (Timmer et al, 7

8 2015) and the OECD Inter-Country Input Output Database (OECD, 2015), among others, are steps towards a solution of the lack of credible IO data. Still, there exists an additional issue when considering the adoption of a CAP approach. Under CAP, consumption is recognised as the main driver of emissions. It is assumed then that the consumer is aware of the environmental impact of the activities that support its consumption and has some control over the technologies used, whereas commonly consumers decisions are driven by the cost of the goods and services either being unaware or disregarding the environmental consequences. Furthermore, consumers, more often than not, do not have any actual control or means of influencing the technology used by the producers. As such, adopting a CAP approach could lead in penalising the consumers that do not have the means to acquire information on the impact of their consumption or the financial resources to switch to cleaner but more expensive substitutes to meet their needs. Our proposed approach recognises that there is merit to both PAP and CAP. A sector that directly generates a large volume of emissions and/or has a large CO 2 footprint merits additional attention. However, the disaggregation of the supply chains allows us to investigate which components of that sector s supply chains hold the most significant share of the total direct emissions or contribute the most on the sector s footprint. Emphasising and analysing these components, that we define as Type (c) hot-spots 3, identifies the role of the requirements of the consumer and the technology of the producer in driving the embodied emissions and which of them should be regarded as the main driver. From a policy perspective, the additional information on the drivers of emissions can be translated into more options for policy makers. For example, if it is found that the requirements of Public Administration from the Electricity sector have high volume of embodied emissions, the proposed methodology/analytical tool can show policy makers the contribution of final demand for Public Administration services, requirements of Electricity output to support Public Administration final demand and the emissions intensity of the Electricity sector (i.e. Electricity production technology) in 3 Detailed definitions of all the different types of hot-spots can be found throughout the thesis starting from Chapter 2 8

9 the total volume of embodied emissions. Using this information, coupled with analyses from other IO analytical tools on the impact on employment and generated value-added, policy makers can decide where policy interventions should take place. The ultimate goal of applying our proposed methodology with other types of IO analysis is to be in a position to allocate the available resources in a more efficient way, achieving significant reduction of emissions, while minimising any potential losses in employment or generated value-added. One of the key features of the proposed methodology is that can be applied both in single region and inter-regional frameworks. For the single region, as it will be discussed in Chapter 2, the analysis can be used in goal setting scenarios. This way the findings can be used as guidelines for policy makers in order to achieve the goals set under international climate change agreements. Of equal importance though is the applicability of the method in inter-regional frameworks. As it was mentioned above, in modern economies, the supply chains are rarely confined within a single country. This being the case, when there are policy efforts in reducing the embodied emissions in either the downstream or the upstream supply chain of any sector in a country, inevitably some components of those supply chains that hold a significant share of the total embodied emissions will be located outside the country. Due to jurisdictional issues quite often it is not possible to impose interventions on industrial sectors outside the territorial borders of a country. However, identifying those hot-spots in the international part of the supply chains could facilitate the reduction of emissions by informing bilateral technology exchange agreements. Even when technology exchange/bilateral co-operation is not possible, since the year 2000 there have been several carbon initiatives promoting the adoption of cleaner technologies in developing nations. These initiatives are referred to using the umbrella term carbon finance and many of them are supported by large international organisations like the World Bank. The results of our proposed interregional analysis then could be used as a basis for determining the funding contribution of developed countries to these initiatives, which in turn would fund emissions-reduction projects. 9

10 1.2 Structure of thesis As mentioned in the previous section this thesis focuses on decomposing the downstream and upstream supply chains of any sector, both in a single country and internationally. We call the findings of our analysis hot-spots. Hot-spots are defined in three general ways. First, they are defined in terms of sectors with more direct emissions than the others in an economy. Second, they are also defined in terms of sectors with a larger footprint than the others in an economy. Third, hot-spots are identified and considered in terms of specific components of a sector s downstream and upstream supply chain (i.e. decomposing information on the first two types). As indicated throughout the previous section, in this thesis we demonstrate that by adopting a hot-spot analysis approach it is possible to get a better understanding on where the emissions are generated and what are the key drivers of their volume. Therefore, the central proposition of this thesis is that hot-spot analysis can lead to more informed and more targeted policy decisions than the headline PAP and CAP analyses that have become so prevalent in the academic and wider research and policy literatures. This thesis is organised as a collection of three papers/chapters, each written as a self-contained piece including literature review and methodology along with original applied work. The second chapter focuses on how hot-spots are defined and considers how standard CAP/PAP accounting methodology may be developed to identify them in an IO framework. To that end a Single Region Input Output (SRIO) framework is used, developing a Chinese case study extracting data for this country from the OECD Inter-Country Input Output database that is used throughout the thesis. The chapter provides guidance on the way that the SRIO is created and how it is extended to an Environmental SRIO (ESRIO), with a focus on CO 2 emissions. Due to what we argue is the need to have a clearer picture of the structure of emissions in and/or driven by different sectors of the economy (and different types of final consumption demand), the ESRIO in this chapter is used in a range of different ways that provide more information 10

11 and a fuller analysis relative to more traditional IO-based CAP and PAP approaches. Specifically, we use the developed methodology to identify the Chinese sectors with the largest volume of direct emissions and the largest footprint, while studying their downstream and upstream supply chains respectively. Moving forward through the thesis, the third chapter reflects the recognition that for most modern economies the supply chains of any given sector extend beyond the borders of its home country. Therefore, here we move to the development of the global Inter-Regional Input Output (IRIO) framework. Adopting a global IRIO approach facilitates the study of the impact of any one sector in any country on the activity and emissions generation in other countries. The third chapter involves an extension of IRIO to an Environmental IRIO (EIRIO). The features of the conventional EIRIO approach are presented and discussed and previous applications in the literature reviewed. However, the key contribution of this paper is to extend the hot-spot identification methodology of the second chapter and apply it using EIRIO decomposition methods. In the applied analysis we focus on the impacts of UK total final demand and identify sectors outside the UK that directly generate significant emissions in serving this demand. We also study in detail the downstream supply chain of the top direct emitting sector outside the UK that produces to support the UK final demand of other sectors globally. Additionally, we identify the sectors with the largest global footprints driven by UK final demands, focusing on UK Health and Social Work activity as a rather interesting example where we can pinpoint important hot-spots in this sector s upstream supply chain components that are located outside the UK. Finally, but of particular importance to the applied contribution made in this thesis, the third chapter is related to an appendix detailing the development of the CO 2 satellite account for the OECD Inter-Country IO (ICIO) database. The author worked towards the development of this dataset in the early stages of the PhD study, during and following a research visit with the ICIO team at the OECD Directorate for Science, Technology and Innovation. 11

12 The fourth chapter goes on to focus on the implications and resolution of a commonly observed tradeoff between IRIO and SRIO, where the latter is based on use of national/regional IO tables as a component of published official statistics. The nature of the trade-off is that IRIO provides crucial information in IO format on multi-lateral international trade but at the cost of industry level detail. On the other hand, SRIO analyses based on national/regional IO tables tend to benefit from a greater level of sectoral disaggregation. Where sub-national regional IO data are available, SRIO also benefits from greater spatial focus, but at the cost of (even) more information loss on intra- and inter-national trade flows. We explore how benefits of both approaches may be maximised, by considering the results of Inter-Regional hot-spot analyses for the UK, focusing on sectors that we then turn our attention to at the Scottish level using more sectorally detailed regional IO data. In doing so, we demonstrate how the information acquired for the case study of UK hot-spots using IRIO and the OECD Inter-Country Input Output database can be enhanced in terms of regional level details by using SRIO and Scottish regional IO tables. Specifically, we demonstrate how the combined UK Electricity, Gas and Water Supply sector, identified as a hot-spot at the Inter-Country database level, may be disaggregated at national and regional level, with identification of further hot-spot activity therein. However, in considering the nature of UK hot-spots at a Scottish level, we also demonstrate the importance of introducing region specific emissions data to give policy makers a more accurate picture of the level of emissions within Scotland. Continuing the focus on Scottish electricity generation within the wider UK sector, we then consider how our results are impacted by introducing data on emissions reported to Scottish Environment Protection Agency by Scottish firms, in place of the UK level average emissions intensity data point provided by the OECD ICIO satellite account. We find that this one impacts the magnitude and ranking of the Scottish sectors with the largest domestic CO 2 footprints. This result highlights the importance of region specific emissions data. It proves that applying generic emissions data, that were created for and initially applied to aggregated sectors at national level, fails to capture the differences in production technologies used in different regions. As a consequence of this any results that are obtained can be over or underestimated, especially in the case of sectors that a multitude of production 12

13 technologies can be used to produce the same output, but with significantly different environmental impact. Therefore, when these results are used to inform policy decisions they could mislead policy makers into either implementing interventions for hot-spots that in reality do not have as many embodied emissions as it was calculated or fail to identify hot-spots that could be rather interesting for policy makers in terms of achieving emissions reduction with low associated costs. The last chapter of this thesis provides a conclusion of this research project as a whole. It summarises the methodology used throughout this project, while reflecting on the strong points and highlighting the weaknesses and challenges that have been discussed in the previous chapters. At the same time, it provides a reflection on the potential future research that can originate from this project. 13

14 Chapter 2: Identifying CO 2 hot-spots in Input Output frameworks: A Chinese single region example. Abstract The chapter uses data for China in 2005 in a Single Region Input Output framework to develop a methodology to identify CO 2 hot-spots in domestic supply chains. It is argued that hot-spot methodology can provide a better understanding of the drivers of sectoral emissions, while allowing a decomposition of the sectors supply chains, thereby offering a clearer view on the structure of emissions. Combining the better understanding of the emissions structure with familiar IO analyses, on employment and value-added for example, can lead to policies that are targeted on the hot-spots where they can achieve significant emissions reduction with the minimum resources cost and losses in employment and value-added. Analysis of the IO data showed that a combined utility sector, Electricity, Gas and Water Supply, is the most polluting Chinese sector in terms of direct CO 2 emissions, whereas Construction has the largest domestic CO 2 footprint, which is mainly driven by Gross Fixed Capital Formation. This finding can be associated with the status of China as a developing country, building up production capacity to facilitate its economic growth. Furthermore, the Construction requirements drive the emissions embodied in the largest hot-spot on Electricity, Gas and Water Supply downstream supply chain, while it is revealed that Electricity, Gas and Water Supply emissions are mainly generated to support the production required to meet the final demand of other sectors. 2.1 Introduction UNFCCC agreements in Kyoto (1997) and Copenhagen (2008) signify important steps in a global effort to mitigate climate change. UNFCCC adopted a territorial Production Accounting Principle (PAP) under 14

15 which every member party, that ratified the agreements, is being held responsible for the emissions generated within its borders (United Nations, 1992). However, even before the implementation of the Kyoto Protocol, the effectiveness of such an approach was doubted. For example, Wyckoff and Roop (1994) suggested that, due to the amount of emissions embodied in imported goods, an approach relying exclusively on domestic emissions for any one country would be less effective in mitigating the emissions of greenhouse gases at the global level. Early indications show that the policy focus will be the same in the upcoming UNFCCC COP21/CMP11 meeting that will be held in Paris in late This means that, at least for the foreseeable future, any international climate change mitigation efforts are likely to be based on a territorial PAP approach. Under PAP a country, or region, is held responsible for the emissions of industries operating within its borders. Numerous studies have conducted comparisons between PAP and a Consumption Accounting Principle (CAP). Under CAP a country s emissions balance includes the emissions generated within its borders in producing goods to be consumed domestically, either as intermediate or final goods, plus the emissions embodied in imports. Thus any emissions generated to produce exported goods, intermediate or final, are not allocated to the country where pollution generation actually takes place rather than where the final consumption driving them is located. The pollution generation aspect of the environmental extension of Input Output (IO) framework has been a commonly used method in studying PAP and CAP. It allows the researchers to study the contribution of different industrial sectors to the total emissions balance of any region/country. Furthermore, it is possible to focus on different types of final demand and study the effect they have on a country s emissions balance. Single Region Input Output (SRIO) has been used in earlier studies (e.g. Munksgaard and Pedersen, 2001; Machado et al, 2001). SRIO is relatively easy to apply since the data required can be extracted by national or regional statistics, which are readily available; often countries publish IO tables. Additionally, there tends to be a high level of details in terms of the types of domestic final consumers identified in regional IO tables. On the down side national IO tables generally lack in details on the use of the exported part of the production. Multi- 15

16 Regional Input Output (MRIO) has been increasingly used in more recent publications (e.g. Munoz and Steininger, 2010; Serrano and Dietzenbacher, 2010), on the grounds that it provides better understanding of the impacts of international trade. However, this often comes at the expense of details on types of final demand. The problem is that the data used for MRIO need to be harmonized, so that they are presented in a common way for all the regions included in the framework. This means that it is often necessary to aggregate industrial sectors as well as types of final consumers, which reduces the level of detail. The common ground between PAP and CAP is that responsibility for the generated emissions is placed on one side or the other. Under PAP it is assumed that the producer is solely responsible for the emissions, which may be appropriate in terms of production technology. Under CAP on the other hand the final demand for goods and thus the end consumer is highlighted as responsible for the generated emissions of any and all industries, according to the perspective that without any demand there would not be any production and therefore any generation of emissions. The aforementioned studies (Machado et al, 2001; Munksgaard and Pedersen, 2001; Munoz and Steininger, 2010; Serrano and Dietzenbacher, 2010) mainly examine the differences in emissions balances between a PAP and a CAP approach. There has been work though, using the IO framework, on the concept of joint responsibility (e.g. Lenzen et al, 2007; Cadarso et al, 2012) essentially recognizing varying degrees of responsibility along the supply chain of each commodity. Wiedmann et al (2007) along with (Wiedmann, 2009) offer detailed reviews on the use of both SRIO and MRIO frameworks, while Wiedmann et al (2011) focus on the evolution of MRIO. 2.2 Hot-spots and key sectors In this chapter, an alternative approach is considered. This is a hot-spot approach that involves decomposition of PAP and CAP but which has received limited attention. In general, a hot-spot can be 16

17 a sector that directly generates a larger amount of emissions or one that its upstream supply chain embodies more emissions than the upstream supply chains of other sectors in a specific region, country or even globally. Alternatively, a point on a sector s downstream or upstream supply chain that embodies emissions above some acceptable level may also be considered as a hot-spot. There are examples of studies identifying hot-spots but either focus on specific commodities or use different economic tools or even study different types of environmental effects. For instance, Acquaye et al (2011) focus specifically on the hot-spots along the biodiesel supply chain, while Turner et al (2012) determined hot-spots in metal manufacture within the Welsh economy (in performing a CGE analysis) and Court et al (2015) field of interest is hazardous waste in domestic supply chains for a range of different types of production and consumption. However, methodologies to help distinguish which sectors and coefficients in an IO framework are the most important in an economy have been developed for many years and there exist studies that discuss on methods to identify those sectors and coefficients. The methods identifying important sectors are usually referred to as key sector analysis and they are applicable at inter-regional, national and subnational level. Rasmussen (1957), Chenery and Watanabe (1958), Hirschman (1958), Dietzenbacher (1992), Sonis et al (2000), Miller and Lahr (2001), Midmore et al (2006) are only some examples of studies that present and discuss on methodologies to identify key sectors. All of them provide different approaches that can be used to identify sectors which have strong forward and/or backward linkages. Our methodology builds primarily on the more classic methods (Rasmussen, 1957; Chenery and Watanabe, 1958; Hirschman, 1958) rather than the eigenvector method (Dietzenbacher, 1992; Midmore et al, 2006). The classic key sector analysis uses the Type I output multiplier as an indicator of a sector s backward linkages. In a similar way we examine each sector s Type I emissions multiplier against its final demand to gauge whether the sector is heavily dependent on polluting inputs or it is the volume of final demand that mainly drives that sector s footprint. In essence, our approach considers the backward linkages of each sector but introduces measurement of emissions to the calculation so that the 17

18 backward linkages are examined from the perspective of the environmental impact. However, our proposed methodology moves forward by disaggregating the supply chains, and therefore the forward and backward linkages, in order to study which of their components are the most polluting. In that sense, our methodology shares a somewhat similar reasoning to what is referred to as important coefficient analysis. Perhaps the most straightforward approach was implemented by Okamoto (2005) who used the value of the average transaction on the 2000 China Multi-Regional IO data (CMRIO) to distinguish the important transactions. However, as Miller and Blair (2009, pp ) describe, there are a number of developed methodologies that identify coefficients in the Input Output coefficients matrix that if they undergo changes they lead to significant changes in the Leontief inverse 4. Even though our approach is different in that we apply our analysis after the calculation of the Leontief inverse, still the two methods have a common motivation; to highlight those elements, of the Emissions multipliers matrix and the CO 2 emissions matrix in our method or the Input Output coefficients matrix in important coefficient analysis, which have a more significant role to play in meeting our different goals set. The goal of this chapter then is to develop a methodology of identifying CO 2 hot-spots in down and upstream supply chains using IO accounting frameworks. The fact that there is no focus on a specific commodity enables the analysis of the data without the need for specialised procedures like Life cycle analysis as observed by Wiebe et al (2012). Moreover, by analysing the observed hot-spots, it is possible to distinguish between those where the absolute amount of emissions is mainly driven by the volume of production and/or those that come as a result of the emissions intensity of the production technology used. Furthermore, while IO is not a sophisticated modelling framework, analysis on the use of multipliers does permit the identification of points in supply chains and/or sectors as a whole, that have the potential to become hot-spots in the event of increased final demand. Below, the third section of this chapter presents the methodology developed here for SRIO multiplier analysis to detect 4 More details on the Leontief inverse, as well as all the matrices mentioned in this section, are presented in the next section. 18

19 CO 2 hot-spots in domestic supply chains, along with a brief summary of the data used. The fourth section then discusses some interesting results that emerge from the analysis while Section 5 concludes and offers suggestions for future research. 2.3 Methodology In this section the adopted methodology is presented. The section starts with a brief presentation of the SRIO providing some details on the way that it has been used in this paper and how a SRIO is expanded to an Environmental SRIO. Then the hot-spot identification methodology is explained through a relatively simple but illustrative example and finally information on the data used in this chapter are provided Single Region Input Output In this chapter the methodology used is based on IO analysis as discussed by Miller and Blair (2009). The key equation of the IO framework used in this study is X = (I A) )* D, (1) where X = X * X 0 X 1 represents the output of sectors i = 1,,N in a N 1 vector format, where X 0 is the total output of sector i. A is a N N matrix, whose elements a 07 denote Input-Output coefficients and represent the input requirements of sector j from sector i = 1,,N to produce one unit of x 7. Finally D, = y * is the diagonal matrix of final demand. The elements of D, may represent total final 0 0 y 1 demand for each sector, but (1) can also be studied in terms of any particular type of final demand such as private (household) consumption, gross capital formation etc. and consider the total output X, driven 19

20 by that final demand source, depending on the focus of study. This approach allows for a focus on one specific type of final demand each time and, when we extend to environmental IO below, to explicitly study the emissions generated in producing output to meet this specific source of final demand. In the case of single region, final demand includes the final goods produced for domestic consumption by different types of domestic final consumers, exported final goods to be used by different types of final consumers abroad as well as exported intermediate goods. In published SRIO for a given country, all exported goods are usually aggregated into a single category including both intermediate and final goods. However, the Inter-Country Input Output (ICIO) table that is used to create the SRIO for our empirical study distinguishes exports to final and intermediate demand in different countries. It thereby, offers the benefit of being able to study the impact of producing intermediate goods in one country to be used as inputs by any sector of any other country included in the ICIO table. For example, it is possible to separately examine the impact of UK s intermediate and private and public final consumption to China by respectively using the data for UK s imports to producing sectors and final consumption sectors in turn as the elements of D, in (1). In moving to the IO multiplier matrix that underlines the model in (1), (I A) )* is an N N matrix that is called Leontief inverse, L. L = (I A) )* = l ** l *7 l *1 l 0* l 07 l 01 l 1* l 17 l 11 (2) The elements l 07 of the Leontief inverse indicate the production required from sector i to meet a unit of final demand of sector j. Post-multiplying the final demand diagonal matrix (for total or any given type of final demand) to the Leontief inverse gives us the following result: LD, = l ** y * l *7 y 7 l *1 y 1 l 0* y * l 1* y * l 07 y 7 l 17 y 7 l 01 y 1 l 11 y 1 (3) 20

21 where y 7 represent the elements of D, along the main diagonal and therefore the final demand for output of sector j. Each element l 07 y 7 then demonstrates the production required from sector i to meet the final demand of sector j. The key difference between the approach used in this chapter and the most commonly used one lies on the presentation of the elements of (3). In most SRIOs, final demand is presented by a N 1 vector or by a N Z matrix that includes all the different types of final demand for each sector. Both these approaches generate aggregated results. In the more general case of postmultiplying L with the N Z final demand matrix, the result is a N Z output matrix that shows the output of each sector that is driven by each type of final demand, i.e. the total output of sector i due to private final consumption or government consumption etc. of all sectors j. On the other hand, in (3) it is possible to examine the output of any sector i due to any selected type of final demand of any sector j, e.g. the output of sector i = Mining and Quarrying required to meet the demand for sector j = Construction gross capital formation. This process requires that each time a different type of final demand needs to be used in D,, but at the same time provides a better understanding of how changes in every type of final demand of each sector are spread throughout the economy. Studying the LD, matrix in (3) allows us to distinguish the different components of the downstream and upstream supply chains of each sector. Each element along a row of the LD, matrix demonstrates the value of the output produced by producing sector i, that is directly required to meet the final demand of sector j, i.e. the downstream supply chain of sector i. In a similar way the columns of LD, matrix represent the upstream supply chain of each sector j, as the elements down each column are the direct requirements of sector j for output from sectors i,so that it meets its final demand. Having this information available enables us later on to identify which of these points on each sector s downstream and upstream supply chains embody a large enough volume of CO 2 emissions to classify them as hotspots. 21

22 2.3.2 Environmental Input Output The goal of this chapter is to create a version of the IO framework that would allow the identification of CO 2 emissions hot-spots. As Leontief (1970) highlights, conventional statistics only report transactions that have some market value. Since environmental impact is generally a non-market transaction (an externality), it is usually omitted and the extent of it should be calculated indirectly. For the purposes of this chapter it is necessary to use CO 2 multipliers. The use of CO 2 multipliers in (1) leads to a transformation of the standard SRIO in a way that the output is expressed in terms of CO 2 emissions rather than monetary value. Extension to environmental IO requires availability of satellite emissions data 5 at sector level, specifically examining total emissions per sector, aggregated at the same level as in the IO tables available. Following that, by dividing the total emissions of each industry group by its total output, the result is what is called an output-emissions coefficient e 0 for each producing sector. Here we focus on a single pollutant, CO 2 emissions, therefore the output-emissions coefficient can also be called CO 2 intensity. What CO 2 intensity represents is the amount of emissions directly generated per monetary unit of output produced. By arranging the emissions coefficients along the main diagonal of a diagonal matrix, the result is the N N matrix E E = e * e E 0 (4) 0 0 e 1 To generate output-emissions multipliers and the environmental IO, E matrix is pre-multiplied to the Leontief inverse. This gives us the basic equation for the environmental IO: EX = E(I A) )* D, (5) 5 A satellite emissions account is a dataset that breaks down the emissions of each sector included in the main IO data tables by fuel type, while also listing the non-combustion emissions (e.g. fugitive gases during extraction of coal, emissions due to industrial processes etc.). 22

23 This leads to the Leontief inverse presented in (2) to be transformed as follows. The result is called Emissions multipliers matrix (Emm for short) Emm = E(I A) )* = e * l ** e * l *7 e * l *1 e 0 l 0* e 1 l 1* e 0 l 07 e 1 l 17 e 0 l 01 e 1 l 11 (6) This approach has been suggested by Court et al (2015) for hazardous wastes. In essence what this process achieves is to transform the Leontief inverse in a way that each element of the matrix demonstrates the amount of CO 2 emitted by sector i so that one unit of final demand in sector j is met. The sum of the elements of 6 down a column is the CO 2 emissions Type I multiplier, commonly called output-emissions multiplier, which shows the total CO 2 emissions generated by all the sectors to meet one monetary unit worth of sector j final demand. Following the same procedure as before, by post-multiplying (6) with the matrix of final demand D,, the result is the CO 2 emissions matrix (Cem). This is the core matrix of the Environmental SRIO and the hot-spot analysis will be conducted primarily on Cem. Cem = EmmD, = e * l ** y * e * l *7 y 7 e * l *1 y 1 e 0 l 0* y * e 1 l 1* y * e 0 l 07 y 7 e 1 l 17 y 7 e 0 l 01 y 1 e 1 l 11 y 1 (7) Following up from the discussion on (3) this method distinguishes itself from the commonly used Environmental SRIO. It provides a decomposed picture of the emissions generated by each sector, revealing the embodied emissions in the different components of each sector s downstream and upstream supply chain. Each element e 0 l 07 y 7 represents the emissions generated by sector i in producing output to support the final demand of sector j. As it has already been discussed the goal of this chapter is to develop a methodology to identify hotspots in domestic downstream and upstream supply chains. The different categories of hot-spots 23

24 identified involve sectors with a larger volume of direct emissions and/or a larger CO 2 footprint compared to the other sectors in an economy. However, it is not necessary that the emissions of each sector, direct or footprint, are evenly distributed amongst the different components of the sector s downstream and upstream supply chain. It is then important to highlight those points that have the largest share of each sector s direct emissions and/or the most significant contribution to each sector s footprint. Analysing the elements of (7) enables the identification of hot-spots belonging to each one of the different categories described. The sum of the elements of each row on (7) calculates the direct emissions of each sector i and can therefore be used to identify the largest direct emitters. Moreover, examining the elements along each row, for each sector i, reveals the emissions directly generated in producing output to meet the sector s own final demand but also to support the final demand of other sectors. By doing so is possible to highlight those components, i.e. hot-spots, on each sector s downstream supply chain that have the largest share of the sector s total direct emissions and therefore need most attention, rather than just focusing on the total emissions of each sector. Furthermore, the sum down the column of each sector j is the sector s CO 2 footprint, allowing for the distinction of those sectors with larger footprint compared to the others in an economy. By focusing on the elements down the column of each sector j in (7) it becomes possible to study the emissions embodied in sector j output requirements from other sectors to meet its own final demand. Part of the arguments in favour of a CAP approach is that, quite often, industrial sectors might not generate significant emissions during their production phase while the different points on their upstream supply chain are generating significant CO 2 emissions. By adopting the suggested methodology, it is possible to identify the points on any sector s upstream supply chain that embody the most significant volume of emissions and therefore make the most significant contribution to the sector s footprint. 24

25 2.3.3 Hot-spot detection The different categories of hot-spots identified in this paper were discussed briefly in the previous section. This section presents the methodology developed to identify the different types of hot-spots. For the purposes of this study as a hot-spot is considered (a) A sector that in producing output generates directly significantly more emissions compared to other sectors in an economy either to support total final consumption demand or components thereof; i.e. has a larger sum of its row in the CO 2 emissions matrix (7). (b) A sector where the output produced to meet its final demand (again, either in total or components thereof), directly and/or indirectly, has a larger footprint, i.e. larger sum down its column in (7), compared to other sectors in an economy (c) A point in a sector s downstream or upstream supply chain, an element within (7), that embodies emissions above a set threshold level in serving all or particular type(s) of final consumption demand. Table 1 is a random example of a simple illustrative single region with only 3 sectors but still rather helpful in understanding the methodology used to identify hot-spots. Identifying Type (a) hot-spots is rather straightforward. By summing the elements along the row of a sector it is possible to calculate the direct emissions generated by that sector. In Table 1 such a sector is 25

26 Mining. It can be seen that Mining is by far the sector with the largest sum of its row. Therefore, Mining can be identified as a Type (a) hot-spot. In a similar fashion it is equally straightforward to identify Type (b) hot-spots. Summing down the column of each sector it is possible to calculate the emissions generated throughout each sector s domestic upstream supply chain to meet each sector s final demand, i.e. the sector s domestic CO 2 footprint. In Table 1 Food and Beverage is the sector with the largest footprint. This identifies Food and Beverage as a Type (b) hot-spot and at the same time the significant difference between the sector s direct emissions and CO 2 footprint suggests that there must be points with large volume of embodied emissions in Food and Beverage upstream supply chain. Focusing solely on the total figures though is not enough. More often than not, specific components of a sector s supply chain have the largest share of the sector s total emissions, both in terms of direct emissions and footprint. At the same time the rest of the components have minor or even negligible embodied emissions due to the fact that the underlying supported output is low and/or the emissions intensity is low. In that sense identifying those key points aids in making more targeted policies, instead of generally trying to make specific sectors more environmentally friendly. To identify this type of hotspots it is necessary to establish an emissions threshold level above which every element of (7) will be considered a Type (c) hot-spot. For illustration purposes, in this chapter, that level is set to be the average of the row maximums. Using averages as a criterion to identify important cells in IO data is not an uncommon approach as it was used for instance by Okamoto (2005) on the 2000 China Multi- Regional IO data (CMRIO) to distinguish the important transactions. However, there is flexibility in setting the threshold level and can be quite easily set at a level to match the suggestions that derive from research outputs in the field of environmental science and/or political decisions. Regarding the latter, the approach adopted in this chapter, and this thesis as a whole, is compatible and easily applicable in a goal setting framework. Recent multinational agreements like the UNFCCC agreements 26

27 and the Europe 2020 strategy, require that the participating parties reduce their greenhouse gas emissions by a set percentage (different for each party) compared to 1990, which is set as a baseline year. Using the CO 2 emissions inventory for 1990, adjusted for the goal of each country that we are interested in, it is possible to calculate the average embodied emissions in each of the transactions within this country. Setting that as the threshold level on the latest IO data would then identify the intersectoral transactions that require policy attention in order to meet the set goals. Looking at the data in Table 1 there are two points that embody more emissions than the average of the row maximums. These are the portions of Mining output required to support Agriculture final demand and Food and Beverage final demand and they can be classified as Type (c) hot-spots. In general, identifying all these different elements and important sectors, i.e. the different types of hotspots, offers policy makers more options as to where interventions can be implemented. However, reviewing the data in Table 1 also reveals the importance of setting the hot-spot threshold level correctly. For example, it can be seen that the output of Agriculture to support the final demand of Mining is not considered a Type (c) hot-spot when using the row maximum average as a threshold level; even though it is the most polluting output of Agriculture. Setting the threshold level lower, for example at 1,000, would then increase the number of Type (c) hot-spots to 5. The scope of this thesis is not to provide a solid way of setting the threshold level rather than to develop the hot-spot analysis methodology and demonstrate ways in which it could be utilised. Still this simple example of sensitivity analysis shows that when applied in real-life scenarios, careful consideration is required when setting the threshold level in order for the analysis to report those components of the supply chains that truly require increased policy attention. Another important point to note is that the approach described here involves identifying the hot-spots on absolute numbers, that is on the total amount of emissions. The main driver of these emissions is final demand. However, total emissions are directly linked to the emissions multipliers of each sector. It 27

28 is worthwhile to apply hot-spot methodology to Emissions multipliers matrix 6 as well, which is in this case is an alternative application of a simple important coefficient approach (Miller and Blair, 2009, p 568). In doing so it is possible to identify elements that have the potential to become hot-spots on absolute numbers if their final demand or that of a heavily dependent sector increases. Therefore, acting proactively, potential sources of significant pollution could be avoided. It is important to highlight though that since IO is not a sophisticated modelling tool, if hot-spot analysis of 6 is employed to draw conclusions on potential sources of significant pollution then the analysis results will have the underlying assumption that the production technology remains exactly the same Data For this study the IO table used is the 2005 Chinese component of the OECD Inter-Country Input-Output Database (OECD, 2015) 6. The database consists of 57 countries, both OECD and non-oecd members, plus the Rest of the World. The version of the database used in this paper shares the same sector grouping as the World Input Output Database (WIOD) (Timmer et al, 2015). The industries of each country have been grouped into 35 sectors following International Standard Industrial Classification of all economic activities (ISIC) rev3.1 (see Appendix 1.A for a complete list of sector grouping). The database also includes: Taxes less subsidies on products Cost, insurance and freight price/free on board price adjustments on exports Direct purchases abroad by residents (imports to final consumption) Purchases on the domestic territory by non-residents Value-added at basic prices International transport margins. 6 This chapter uses an earlier pilot version of the database. We would like to express our gratitude to OECD Directorate for Science, Technology and Innovation for sharing the database with us and for their continuous support throughout this research project. 28

29 On the final demand side there are five types of final demand included: Private (Household) Consumption Non-Profit Institutions Serving Households Government Final Consumption Gross Fixed Capital Formation Inventory (changes in stocks). It is important to note at this point that the data are being reported in monetary value of the goods and services traded. China has been selected for this single region study because it is generally recognised as a major emitter of CO 2 and also a major exporter. The satellite emissions data used come from the WIOD project (Genty, 2012). Among the satellite accounts created from that project there is a detailed table with the emissions generated by each sector, derived from 26 combustible and non-combustible energy sources as well a non-energy category. The sectoral classification matches the one of the OECD database used in this chapter; therefore, there is no need to use a different sector mapping. The WIOD emissions table also includes the emissions of the households as an aggregated category. Household direct emissions are generated as a direct result of consumption of final goods, but they are not part of a sector s downstream supply chain. However, if it is required to calculate the total emissions generated due to private consumption then the household direct emissions should be included in the calculations. 2.4 Results of hot-spot analysis In this section some interesting results from the analysis are presented. The hot-spot analysis was conducted on CO 2 emissions matrix (7), which was calculated using total final demand. The first subsection focuses on the direct emissions of Chinese sectors. Electricity, Gas and Water Supply is found to 29

30 be the largest Type (a) hot-spot and there is additional focus on its downstream supply chain to identify potential Type (c) hot-spots. In the second sub-section the point of focus is the footprint of Chinese sectors. The largest Type (b) hot-spot, i.e. the sector with the largest footprint, is found to be Construction. Almost the entire output of Construction is produced to meet its final demand, which is also the largest amongst Chinese sectors, so the sector s upstream supply is examined to look for Type (c) hot-spots Direct emitters As stated in the previous paragraph the first point of focus of this chapter are the direct emissions of the Chinese sectors. Table 2 demonstrates the top 10 most polluting sectors in terms of direct emissions, i.e. the sum of the elements along each sector s row in (7). In Table 2 the first column of results ( Direct Emissions ) refers to the sum of the elements along each sector s row in (7). In the second column these emissions are presented as a percentage share of the total direct emissions. The third column includes each sector s e 0 element of matrix E (4), while the fourth column shows the y 7 element of each sector in the final demand matrix D,. Finally, the last column is the sum of the elements along each sector s row in (3). 30

31 Table 2. Top 10 direct emission producers (row totals in (7)) Rank WIOD Sector Code Sector Direct Emissions (kt of CO2) % share of Total Direct Emissions CO2 Emissions Intensity (kt of CO2/$m of Total Output) (in (4)) Total Final Demand ($m) Total Output ($m) 1 E Electricity, Gas and Water Supply 2,465, % , , Other Non-Metallic Mineral 534, % , , t28 Basic Metals and Fabricated Metal 507, % , , Chemicals and Chemical Products 198, % , ,557 5 C Mining and Quarrying 135, % , ,276 6 AtB Agriculture, Hunting, Forestry and Fishing 134, % , , Inland Transport 85, % , , Coke, Refined Petroleum and Nuclear Fuel 81, % , , Water Transport 81, % ,967 66, F Construction 57, % , ,417 All other sectors 403, % Total Direct Emissions 4,685, % As can be seen in Table 2, sector Electricity, Gas and Water Supply (hereafter EGWS) is the largest direct emitter of CO 2 in China. The emissions generated by EGWS significantly outweigh the emissions of any other sector. EGWS alone generates more CO 2 than all the remaining sectors combined. Apart from EGWS, sectors Other Non-Metallic Mineral and Basic Metals and Fabricated Metal generate significant emissions. These 3 sectors are the top 3 direct polluters in China and can be identified as Type (a) hot-spots. In fact, the direct emissions of EGWS, Other Non-Metallic Mineral and Basic Metals and Fabricated Metal combined, account for 74.85% of the total Chinese emissions. It is important though to explore what is the main driver of the direct emissions. By examining the data in Table 2 it is easily noticeable that the difference in generated emissions between EGWS and the remaining sectors is due to EGWS CO 2 emissions intensity, which is the largest in China. EGWS is more than 3.5 times more emissions intensive than the second most CO 2 intensive sector, Other Non-Metallic Mineral. These figures are a direct result of China s heavy dependence on fossil fuel to produce electricity, especially in 2005 where major projects like the Three Gorges Dam hydroelectric power plant were not in operation yet. 31

32 Emissions intensity is also the main driver of the total direct emissions of the second largest direct emitter, Other Non-Metallic Mineral. On the other hand, the direct emissions of the third sector of Table 2, Basic Metals and Fabricated Metal, are a result of the sector having the 5th largest emissions intensity and at the same time being the 3rd largest in terms of value of total output. Even though data show that the majority of the direct emissions are generated due to the operation of emissions intensive sectors, there are sectors featured on Table 2, despite their emissions intensity being relatively small. For example, the CO 2 intensity of Agriculture, Hunting, Forestry and Fishing is significantly lower than the one of Water Transport. Still the total emissions of Agriculture, Hunting, Forestry and Fishing are significantly more than the ones of Water Transport. It is the total volume of production that drives Agriculture, Hunting, Forestry and Fishing onto the 6th position of the direct emitters. In fact, Agriculture, Hunting, Forestry and Fishing is ranked 4th in gross total output amongst the 35 Chinese industrial sectors. In any case the amount of emissions generated by Chinese EGWS, classify the sector as the top Type (a) hot-spot. However as seen in Table 2, only 6.66% of the sector s total output is produced to meet the sector s final demand. In fact, out of the sector s total emissions, only 9.66% are generated for its own final demand. This part includes EGWS own requirements to produce for its final consumers. So it is important to look for hot-spots on the EGWS downstream supply chain, which are presented in Table 3. 32

33 In Table 3 the Embodied Emissions column (first column of results) shows the embodied emissions in the requirements of each sector of Table 3 for output from China s EGWS, i.e. the element of each sector on the row of EGWS in (7). In the second column the emissions of the first column are presented as a percentage share of the total direct emissions of EGWS. The third column shows the element of each sector listed in Table 3 on the row of EGWS in (6). The data from this column help understand whether each of the listed sectors has high requirements for output of EGWS to meet its final demand. Finally, the fourth column includes the final demand of each sector, i.e. the y 7 of each sector in final demand matrix D,. For an element of Chinese matrix (7) to be considered as a Type (c) hot-spot it has to be above the average of the row maximums of (7), which is 44,900 kt of CO 2. On EGWS downstream supply chain there are 13 Type (c) hot-spots, the most compared to any other sector. As can be seen from Table 3 the Type (c) hot-spots on EGWS downstream supply chain account for 82.9% of the sector s total direct emissions. The largest share is held by Construction, the requirements of which lead to the generation 33

34 of 22.86% of EGWS emissions. Construction is also the largest Type (b) hot-spot as it will be shown in the following sub-section and its requirements for EGWS output make a significant contribution to the sector s footprint. Examining the emissions multiplier, element of (6), and the final demand of Construction it seems that even though the emissions multiplier is slightly higher than the average of the other Type (c) hot-spots, 1.06 compared to an average of 0.89, it is Construction final demand that leads to significantly more emissions generated. From Table 3 it is noticeable that sectors Basic Metals and Fabricated Metal and Chemicals and Chemical products have significantly higher requirements for EGWS output, almost twice as much as the average of the Type (c) hot-spots. Therefore, even though the value of their final demand is amongst the lowest on Table 3, still their requirements led to more CO 2 to be emitted, especially Basic Metals and Fabricated Metal, compared to other sectors with significantly higher final demand. However, the more characteristic example are the requirements of EGWS to meet its own final demand. The emissions shown in Table 3 though include both the CO 2 emitted during the production of output for the sector s final consumers and the emissions generated to support the production of that output. Out of the 238,092 kt of CO 2 emitted by EGWS to meet its own final demand, 164,304 kt are generated in producing final goods (i.e. the product of the sector s final demand multiplied by the CO 2 emissions intensity) while the remaining 73,788 kt are generated to support the production of output to meet EGWS final demand. The problem that arises in this case is that EGWS is an aggregated sector 7. It is not possible then, using the data from the IO table, to make any distinction on which part of the 73,788 kt are generated by Gas to support the production of Electricity or by Electricity to support Water Supply. In the same way it is not possible to identify what is the share of each sector included in EGWS in producing output to meet EGWS final demand. 7 Apart from the inability to make a distinction between the contribution of the different sectors aggregated into EGWS, there are problems that arise from the aggregation itself. Chapter 4 provides a more in-depth discussion on the aggregation problem. 34

35 2.4.2 Emissions attributed to final demand The data of Table 2 show that for the majority of the sectors featured in Table 2 there is a significant difference between their final demand and their total output. The implication is that these sectors mainly produce output to support the final demand of other sectors. That being the case they must be featured on the upstream supply chain of other sectors that they might not be featured in Table 2. For instance, it can be seen from Table 3 that the downstream supply chain of EGWS includes Type (c) hotspots in sectors that are not in Table 2. It is necessary then to determine which Chinese sectors can be classified as Type (b) hot-spots, i.e. have the largest footprint, and focusing on the largest one, whether there are any Type (c) hot-spots in its upstream supply chain. Table 4 shows the top 10 Chinese sectors in terms of domestic footprint. The domestic footprint of each sector is the sum of the elements of its column in (7). The first column of Table 4 shows the footprint of each sector, i.e. the sum of the elements down the column of each sector in CO 2 emissions matrix (7). In the second column the footprint of each sector is presented as a percentage share of the total footprint of the Chinese sectors. Then in the next three columns the footprint is broken down by source of final demand. In the third column the focus is on private (household) and government consumption, in the fourth column on gross fixed capital formation and in the fifth column on exports. The numbers in columns 3, 4 and 5 are the sum of the elements down the column of each sector in (7), when 7 was calculated using private and government consumption, gross fixed capital formation and exports respectively as the elements of the final demand matrix D,. 35

36 36

37 A first look in Table 4 reveals that some of the sectors included have a significant difference in their ranking in terms of footprint compared to their ranking in terms of direct emissions. Especially in the case of the top two sectors, Construction and Electrical and Optical Equipment, the emissions generated to meet their final demand is over 22 and 38 times higher respectively than their direct emissions (the direct emissions of Electrical and Optical Equipment are 13,111 kt of CO 2 ). This suggests that the magnitude of the footprint of both Construction and Electrical and Optical Equipment is due to the requirements from other sectors rather than Construction and Electrical and Optical Equipment own production. Therefore, Construction and Electrical and Optical Equipment are the largest Type (b) hot-spots. A rather interesting observation comes by substituting total final demand in (7) with different sources of final demand. By doing so it is possible to capture the impact of a specific type or groups of different types of final demand and the results can be seen in the third, fourth and fifth column of Table 4. For Chinese Construction this shows that the majority of the sector s footprint is associated with domestic gross fixed capital formation, which is the main type of final demand for this sector and linked to development of production capacity to enable economic development. This observation is line with China s status as a developing country. For Electrical and Optical Equipment the main type of final demand, and therefore the one that supports the majority of the sector s footprint, is exported final demand. Electrical and Optical Equipment mainly produces output to be exported as final goods, 55.9% of the sector s exports, or intermediate goods, 44.1% of sector s exports. In total, Electrical and Optical Equipment has the second highest total final demand in China in 2005, 69.5% of which is exported. In a similar fashion the footprint of Basic Metals and Fabricated Metal and Textiles and Textile Products is also supported by exports. Having 3 exporting sectors, Electrical and Optical Equipment, Basic Metals and Fabricated Metal and Textiles and Textile Products, in the top 10 in terms of footprint seems somewhat logical since China is considered to be one of the major exporters. 37

38 Contrary to Construction and Electrical and Optical Equipment, two of the most polluting sectors in terms of direct emissions (see Table 2), EGWS and Basic Metals and Fabricated Metal, are located significantly lower on the list of the Chinese Type (b) hot-spots (see Table 4). In the case of EGWS the difference between the direct emissions (sum of EGWS row in (7)) and the emissions generated for the sector s final demand (sum of EGWS column in (7)) is over 2.2 billion tons of CO 2, i.e. over 46.9% of the total emissions generated by Chinese industries. Another interesting observation is that, with the exception of the top two sectors in terms of footprint, the emissions are spread more evenly across the whole of the remaining sectors. In essence the majority of the sectors have somewhat even shares of the total emissions attributed to final demand compared to the direct emissions case, where a handful of sectors had the most significant impact and the rest had only minor shares of the total direct emissions. As in the case of direct emissions it is important to be in a position to determine the main driver of the footprint of the top 10 sectors of Table 4. Table 5 shows the Type I emissions multipliers, i.e. sum down the column of each sector in (6), for the top 10 sectors with the largest footprint. 38

39 As in Table 4 the Total Emissions column is the sum of each sector s column in (7). The Type I Emissions Multiplier column refers to the sum of each sector s column in (6) and the Total Final Demand column has the y 7 element of final demand matrix D, for each sector of Table 5. Analysis of the data shows that in China the average Type I emissions multiplier is 1.93 kt of CO 2 /$m worth of final demand. Examining the figures in Table 5 it can be seen that the majority of the top 10 sectors in terms of footprint have a Type I emissions multiplier quite close to the average. It is clear then that the magnitude of the Type (b) hot-spots is mainly driven by the volume of final demand, rather than the Type I emissions multiplier. In fact, 8 out of the 10 sectors of Tables 4 and 5, are the ones with the highest final demand in China. The only two exceptions are EGWS and Basic Metals and Fabricated Metal that even though their final demand is significantly smaller than the rest of the sectors in Table 5, their Type I emissions multiplier is much larger, causing them to be featured on the table. For Construction, the largest Type (b) hot-spot in China, the result is a mixture of the fact that it has the 7th largest Type I emissions multiplier and the largest volume of total final demand amongst all Chinese sectors. However, as previously observed the difference between direct emissions and footprint suggests that the major contribution to the sector s footprint is made by the sector s requirements for other sectors output. The implication of the observations on Construction is that there must be several Type (c) hot-spots on the sector s upstream supply chain. These hot-spots should make the most significant contribution to Construction footprint. Therefore, it is important to study the structure of Construction upstream supply chain. Table 6 shows the Type (c) hot-spots along with row maximums on Construction upstream supply chain. The first column of Table 6 shows the elements of each sector on the column of Construction in (7) while the second column presents the elements as a share of the total emissions in the Type (b) hot- 39

40 spot. The Emissions Multiplier column has the elements of each sector on the column of Construction in (6) while the Output Multiplier are the elements of each sector on the column of Construction in Leontief inverse (2). The benefit of having both the Emissions Multiplier and the Output Multiplier column is that it offers a clearer view on whether the embodied emissions (first column on Table 6) are a result of Construction requiring a large volume of output from a specific sector or that sector s production is CO 2 intensive. As a reminder the threshold level above which a point of (7) is considered a Type (c) hot-spot is 44,900 kt of CO 2. As seen in Table 6 there are 4 Type (c) hot-spots on Construction upstream supply chain. At the same time there are also the row maximums of 5 more sectors, on Construction upstream supply chain, which do not embody enough emissions to be classified as Type (c) hot-spots. The results reported in Table 6 verify that the magnitude of Construction Type (b) hot-spot is mainly due to the emissions generated by other sectors to support Construction final demand. Construction own production to meet its final demand contributes 4.32% to the sector s footprint. The requirements for output from EGWS as intermediate input of Construction make the most significant contribution, followed by the output from Other Non-Metallic Mineral and Basic Metals and Fabricated Metal. Interestingly these three sectors, EGWS, Other Non-Metallic Mineral and Basic Metals and Fabricated Metal, are also the ones with the highest amount of direct emissions as seen on Table 2. Additionally, the output from Other Non-Metallic Mineral to support the final demand of Construction embodies 67.8% of the total direct emissions of Other Non-Metallic Mineral. In fact, all the elements of (7) shown on Table 6 are the row maximums, i.e. the most polluting outputs, of the respective sectors. Interestingly enough, even though the upstream supply chain of Construction includes a number of Type (c) hot-spots and row maximums, the output of Construction is not a Type (c) hot-spot in any other sector s upstream supply chain. This observation is to be expected as the data from Table 2 show that Construction output is mainly final goods to meet the demand of final consumers, rather than intermediate goods for other Chinese sectors. 40

41 41

42 The fact that Construction has the largest volume of final demand amongst the Chinese sectors plays a significant role. At the same time, it is important to distinguish whether the Type (c) hot-spots in Construction upstream supply chain are driven by each sector s CO 2 intensity or Construction high requirements for output. Examining the emissions multipliers column of Table 6 is a first step. It can be seen that the requirements for output from EGWS and Other Non-Metallic Mineral have the larger emissions multiplier, which, due to the large final demand, leads to large amounts of embodied emissions. However, that does not necessarily mean that Construction requires large amounts of input from EGWS and Other Non-Metallic Mineral. Examining the output multiplier column of Table 6, which shows the elements of Leontief inverse (2), it can be seen that in the case of EGWS, Construction require less output per $m of final demand compared to requirements from other sectors. Still the large CO 2 emissions intensity of EGWS seen on Table 2 leads to a significant impact in terms of embodied emissions. Other Non-Metallic Mineral is somewhat similar, being the second most CO 2 intensive sector, but at the same time Construction production is significantly more dependent on output from Other Non-Metallic Mineral than EGWS. In fact, Construction is most dependent on inputs from sectors Other Non-Metallic Mineral and Basic Metals and Fabricated Metal, with different degrees of associated emissions per unit of final demand, due to the different CO 2 emissions intensities. 2.5 Conclusions and extensions Input Output is a well-established methodology to study the emissions generated as a result of economic activities. However, previous studies mainly used aggregated results in terms of Consumption Accounting Principle (CAP) and Production Accounting Principle (PAP) and focused their interpretation in terms of these two headline indicators on how they relate to each other. As a result, there is limited information obtained regarding the structure of the emissions, the generation of which and their drivers appear to be the output of a black box. Without a more detailed picture of the structure of the

43 emissions, policy decisions can only be based on headline figures like the CAP and PAP emissions, which in turn means that policies target whole sectors that are identified as polluting rather than focusing on the components of their supply chains that hold the major shares of the emissions. The goal of this study was to develop a method that provides transparency in the structure of both the direct emissions and the CO 2 footprint of each sector. Furthermore, by using such a transparent method, we established an approach by which it is possible to distinguish the most important amongst the different components of any given supply chain, i.e. the hot-spots. In principal, the hot-spot methodology builds on elements of the established key sector and important coefficient analysis methods. The Type I multipliers presented in Table 5 are the analogous to the backward linkages of each sector in key sector analysis, whereas the discussion on output multipliers of Table 6 is in essence a basic important coefficient analysis. To these elements of key sector and important coefficient analysis, the hot-spot methodology adds the analysis of emissions intensities, emissions multipliers, final demand and total output, so that it generates an as comprehensive report as possible on the structure and the driver of each sector s emissions. We believe that the comprehensive nature of this analysis will be a useful tool for policy makers in developing climate change mitigation policies and meeting the goals set as part of international agreements. Furthermore, the identification of specific components of each sector s supply chains offers the potential for targeted policies. For example, the analysis of the results in this chapter has revealed that Chinese Construction sector has the largest domestic CO 2 footprint amongst all sectors in China. Without a clear picture of the structure of the Construction footprint, the policy option might have been to restructure the sector as a whole. However, hot-spot analysis has revealed that the requirements from the Chinese EGWS sector contribute significantly to the footprint of the Chinese Construction sector. This finding indicates to policy makers that significant gains can be made in terms 43

44 of reduced CO 2 emissions either by reducing the Construction requirements for output from EGWS or by making EGWS less emissions intensive. As mentioned previously in this chapter the reasoning behind identifying hot-spots is to highlight sectors as a whole and components of the supply chains of any sector, which embody significant emissions and therefore merit attention at the policy level. A similar driver led to the development of key sector analysis so that policy makers would be in a position to identify sectors where increased emphasis and allocation of resources would maximise the economy-wide benefits (Miller and Lahr, 2001). However, in this research project the hot-spot methodology is used (and proposed to be used) to provide an ex-post analysis of the embodied emissions in supply chains, rather than for ex-ante planning purposes for which, as Midmore et al (2006) have pointed out, the usefulness of IO frameworks is limited. This is one of the main limitations of the proposed methodology. It is a specialised tool focused on the identification of the most important amongst a number of elements. To be used for planning purposes it would be necessary to combine hot-spot analysis with commonly used IO analysis tools like employment and value-added multiplier analysis so that it can be identified which of the hot-spots present the minimum risk possible for losses in terms of employment positions and value-added generated. Following that the use of a sophisticated modelling tool like Computable General Equilibrium on the selected sectors would provide a more advanced representation of what would be the economywide impact of interventions on the selected hot-spots. The other main limitation is that hot-spot analysis is a static picture of an otherwise dynamic system. This problem could be rectified by conducting a historical hot-spot analysis in multiple years, which would reveal the evolution of the hot-spots due to economic changes and/or implemented policies. At 44

45 the same time though it would help clarify which of the initially identified hot-spots manifest as ongoing issues and which, if any, are just one-off problems that disappeared or were resolved. However, historical analysis is only one of the potential steps forward in further applying hot-spot analysis. Since the modern economy is globalised, the supply chains rarely are restricted within the borders of a single country. Therefore, it is important to extend the methodology introduced here to the global level and analyse the impact of international trade. Applying this methodology to Inter-Regional Input-Output seems to be a logical next step. IRIO offers the ability to study the supply chains of sectors in multiple countries. Furthermore, as we are not limited to the domestic side of supply chains, it is possible to identify hot-spots on the international part of any supply chain. 45

46 Chapter 3: Carbon hotspots in global supply chains: an inter-regional input-output analysis. Abstract This chapter extends the CO 2 hot-spot identification methodology to an Environmental Inter-Regional Input Output (EIRIO) accounting framework. Drawing on OECD Inter-Country Input Output tables, we decompose standard EIRIO headline calculations of production and consumption perspective emissions. We show how this facilitates consideration of downstream demands driving the production and associated emissions of CO 2 at specific industrial hot-spots outside the borders of individual regions/territories under study. The results for a UK study reveal how domestic final demand can drive the generation of emissions outside the UK s territorial boundaries (despite the fact that the majority of emissions are generated within its borders). The combined Chinese Electricity, Gas and Water Supply sector is identified as a major direct emitter of CO 2 in the global supply chain of other industries (including UK-based ones) serving UK final demands. Furthermore, the UK Health and Social Work sector is revealed to have the second largest CO 2 footprint driven by UK final demand, amongst all production sectors in all countries. However, it is found to have numerous CO 2 hot-spots in its international upstream supply chain, highlighting the impact of UK s Health and Social Work sector on generation of emissions in the UK s trading partners. 3.1 Introduction Over the last two decades the Input Output (IO) framework has become a widely used tool in studies related to the environmental impact of economic activities. One of the most commonly studied fields is 46

47 the structure of CO 2 emissions under different accounting principles. The prevailing policy approach internationally for mitigating climate change, adopted by UNFCCC, assigns responsibility to the participating members only for the emissions generated within their territory. This Production Accounting Principle (PAP) has been met with scepticism by numerous researchers (e.g. Wyckoff and Roop, 1994; Munksgaard and Pedersen, 2001) mainly due to the issue of emissions embodied in international trade, which may be overlooked by an approach that focuses on a single country. Furthermore, as Arrow et al (1995) discuss, developed countries can achieve their emissions reduction by moving their high emissions generating activities abroad, i.e. the phenomenon identified as carbon leakage. The fact that emissions-intensive goods are often produced for exports led to the suggestion that a Consumption Accounting Principle (CAP) would be a better way of assigning responsibility for the generated emissions. Under CAP the ultimate responsibility is assigned to the consumer of any given product or service, assuming that demand for production is the driver for any emissions, and therefore alleviating any responsibility from the producer. However, early indications show that policy focus in the upcoming UNFCCC COP21/CMP11 (in Paris, November 2015) will remain the same, which means that territorial PAP will most likely be the approach used to calculate emissions and assign responsibility for action, at least in the medium term. Additionally, policy makers, in the UK for example, have raised a number of issues associated with the implementation of consumption-based measures. As reflected in a report from the House of Commons Energy and Climate Change Committee (2012a), the Department of Energy and Climate Change (DECC) has highlighted the lack of robust and transparent data on international trade that would be crucial in designing consumption-based measures. Furthermore, the UK government, responding to the aforementioned report, has also brought forward the existence of practical complications due to the product-specific nature of the consumption-based emissions (House of Commons Energy and Climate Change Committee, 2012b). 47

48 These by no means suggest that CAP should be disregarded, especially since there have been steps towards the direction of resolving the lack of robust data by publishing detailed Inter-Regional Input Output (IRIO) tables like the World Input Output Database (Timmer et al, 2015) and the OECD Inter- Country Input Output database (OECD, 2015). In fact, in the same report by the UK House of Commons Energy and Climate Change Committee (2012a), it is highlighted what the potential benefits are from implementing consumption-based policies. However, it seems preferable to find a way to gather information and develop techniques to consider insights from both PAP and CAP measures to make more informed policy decisions. Moreover, demand is not the sole driver of emissions and more often than not consumers do not have any direct control of the production methods used. In this sense CAP analysis may be regarded more as a useful approach in understanding the main economic pressure points that drive the emissions generated by producing sectors rather than a responsibility principle as such. As shown in the previous chapter of this thesis for a Single Region IO (SRIO), a hot-spot approach that studies the emissions from the perspective of both the producer and the consumer offers a better understanding on the drivers and the structure of emissions. However, in a globalised economy, supply chains are not restricted within the borders of a country. It is necessary then to apply the methodology in a more global framework. IRIO provides such an accounting framework. Analysing the data available in the OECD Inter-country Input-Output (ICIO) Database 8 (OECD, 2015) 9 it is possible to identify CO 2 emissions hot-spots in global supply chains, varying from direct emitters and industrial outputs with large overall footprints (in serving final demands) to specific points in sectors supply chains that embody significant volumes and/or intensity of emissions. IRIO ensures that emissions embedded in international trade will be accounted for during the analysis. Therefore, it enables the decomposition of emissions embodied in downstream and upstream flow of goods for any given sector. The next section 8 Other published IRIO datasets include WIOD and GTAP. More details are provided in the following section. 9 In this thesis an earlier, pilot, version of the database is used. We are thankful to the OECD Directorate for Science, Technology and Innovation for providing access to the database and also for all the support in terms of collaboration and exchange of ideas and additional data. Their contribution has been invaluable. 48

49 provides a review of the existing relevant literature. Section 3.3 expands on the methodology and the data used while in Section 3.4 some key results are presented and discussed. The final section concludes and offers suggestions for further research. 3.2 Literature background Early CAP and PAP studies used SRIO (e.g. Munksgaard and Pedersen, 2001; Machado et al, 2001). A SRIO can be created by using national IO tables (generally part of national statistics) along with necessary satellite data for the environmental extension (e.g. emissions per industrial sector, which may or may not be provided through national accounting or other official published data sources). SRIO data provided through national statistics tend to have a greater level of sectoral detail than the Multi- Region IO (MRIO) or IRIO data provided by bodies such as OECD, the WIOD project or GTAP, and often also offer more break down of domestic final consumers (e.g. breaking out tourist demand from household consumption). The downside of SRIO with imports and exports reported in an aggregate row and column respectively - is the lack of information on the emissions impact of international trade. Even where an underlying use matrix is available to identify both domestic and imported goods and services imported to each production sector and final consumer by output or commodity type, this is likely to be aggregated at Rest of World (ROW) category with no information on pollution technologies of the industries located in different countries. This way any analysis on emissions impacts attributed to final demand in SRIO tends to be conducted under the domestic technology assumption that the imported goods have been produced using the same technology as the examined economy (see Turner et al, 2011a). In a globalised economy this could lead to reduced accuracy and credibility of any findings. In an effort to capture the environmental impacts of international trade, there were efforts to produce MRIO frameworks (see Wiedmann et al, 2007; Wiedmann, 2009 for detailed reviews of SRIOs and 49

50 MRIOs). MRIO and IRIO differ in terms of the detail incorporated in matrices recording inter-country transactions. However, the common feature of MRIO and IRIO is that they include inter-country transactions explicitly for every country in the framework, without having the imports and exports in aggregated categories (columns/rows) as in SRIO. Therefore, an approach based on MRIO or IRIO provides a fuller insight on the pollutants emitted to produce goods that will be used either as intermediate or final goods outside the territory of each directly emitting country. In one of the studies using MRIO, Lenzen et al (2004) expanded the work of Munksgaard and Pedersen (2001) to include Denmark and some of its major trading partners (Sweden, Norway and Germany) as well as ROW. Among their findings, Lenzen et al (2004) demonstrated that as they moved towards a scenario where country-level data on production (and polluting) technologies were incorporated, the emissions attributed to each country differed significantly. Denmark, for instance, was proven to be an emissions importer instead of an exporter as calculated in the SRIO analysis of Munksgaard and Pedersen (2001). These findings suggest that, in order to improve the accuracy of our findings, it is of key importance to use frameworks as detailed as possible when it comes to the production technology and trade relationships of the countries included. In fact, the work by Shui and Harriss (2006) on the impact of trade between China and USA demonstrates that trade relations between very large and open economies may have a significant effect on global emissions. In general, MRIO/IRIO have been used for a variety of types of analyses, including estimation of a range of different types of footprints (ecological, carbon and water footprint), as well as materials use embodied in international trade (e.g. Munoz and Steininger, 2010; Serrano and Dietzenbacher, 2010; Bruckner et al, 2012; Ewing et al, 2012). The progress and beneficial characteristics of MRIO/IRIO over the last years has been discussed by Wiedmann et al (2011). This review also provides an insight on what might be the requirements from future researchers who opt to use MRIO/IRIO analyses in determining the environmental impact of human activities. MRIO tends to be used where there are limitations on interregional trade data. Therefore, we hereon refer to the full IRIO approach. 50

51 A common research interest amongst the studies using both SRIO and IRIO has been the allocation of responsibility for the emissions generated and investigating the differences between PAP and CAP findings for given countries under study. However, focusing on the differences in allocated emissions under different accounting principles does not necessarily offer a better understanding on the structure of the emissions. Turner et al (2007) moved towards the direction of a more in-depth study of the IRIO underlying matrices by using the IRIO theory to establish a method that can capture both the direct and indirect effects of human economic activities. Their method calculates the ecological footprint; however, by substituting the resource-use matrix with emission intensities matrix then the model can be used in the carbon footprint framework. In fact, McGregor et al (2008) used this approach to calculate the CO 2 trade balance between Scotland and the rest of UK. A similar approach has been suggested when studying the concept of shared responsibility (e.g. Lenzen et al, 2007; Cadarso et al, 2012) where, due to the need to allocate responsibility to different points along supply chains, it was necessary to decompose the total emissions/footprint figures. One of the most recent IRIO is the Global Resource Accounting Model (GRAM) introduced by Wiebe et al (2012), who use the OECD Inter-Country Input-Output (ICIO) accounts to calculate the emissions embodied in international trade originating from energy use. The method used by Wiebe et al (2012) shares significant similarities to the approach discussed by Turner et al (2007). The papers by Turner et al (2007) and McGregor et al (2008) also highlight a number of issues that need to be addressed in order to generate credible results. The most significant one is the requirement for highly detailed datasets that meet specific characteristics such as: (a) all the transactions between the countries included reported in IO format with (b) common sector classifications and (c) inclusion of direct imports of final goods and detailed imports of intermediate goods. Therefore, IRIO tables are difficult and resource intensive to produce, providing one reason why IRIO has not been extensively used until recently. Amongst the existing IRIO datasets, one of the most extensively used ones is the 51

52 World Input Output Database (WIOD) (Timmer et al, 2015). The WIOD dataset includes 40 countries plus Rest of the World (ROW) with 35 production sectors in each. The data have been harmonized in a way that the table of every country included has the same sector classification and the transactions are reported in US dollars (USD) across the board. Additionally, an array of social and environmental satellite accounts is included to facilitate the use of WIOD in a variety of fields. However, in this thesis the OECD ICIO database is used as it benefits from a larger number of countries (57 plus ROW) and less aggregated sectors in each country (37 sectors rather than 35 in WIOD). The fact that the sectors are grouped differently compared to WIOD also meant that it was necessary to create a satellite emissions account for use in the environmental IRIO, rather than using the one published as part of the WIOD project. 3.3 Methodology and data Inter-Regional Input Output In the previous chapter, hot-spots were identified in a SRIO. However, to study the generated emissions due to international trade it is necessary to use an IO framework that includes multiple regions, or countries as in this study. The basic IO equation of a framework with 2 regions, a simpler version of the framework used in this chapter, is the following. X ** X *E X E* X EE = ( I* 0 0 I E A** A *E Y ** Y *E A E* A EE ))* Y E* EE (1) Y X ** is a N 1 vector of output of every sector i = 1,, N produced and supported by final consumption demand originating in region 1, while X *E is the output produced in region 1 and supported by final consumption demand originating in region 2 (via export demands Y *E ). In the same way X E* is the 52

53 output produced in region 2 and supported by export demand from region 1 while X EE is the output supported by domestic final consumption demand in region 2. Each N N matrix A is called an inputoutput coefficients matrix. For example each element a *E 07 of matrix A *E shows the intermediate purchase of input from sector i in region 1 as a share of total input in sector j output in region 2 (i, j = 1,, N). The key point to note, relative to SRIO, is that the elements of A *E and A E* now become part of (endogenous) intermediate matrix rather than (exogenous) final demand (exports) and primary input (imports). In the framework used in this study the output of each sector is reported in monetary value, in millions of US dollars (USD millions). Finally, Y ** is a N Z vector of final demand for output from the sectors in region 1 by final consumers in region 1, while Y *E is the final demand for output from the sectors in region 1 that is exported to final consumers in region 2. Similarly, Y E* is the final demand for output of sectors in region 2 exported to final consumers in region 1 while Y EE is the domestic final demand for output from the sectors in region 2. Each element y *E 7M of Y *E represents the type z, z = 1,..., Z, final demand for output of sector j in region 1 exported to of final consumers in region 2. Types of final demand include public and private (household and government) final consumption or capital formation. In this way it is possible to identify the output in regions 1 and 2 supported by specific types of final demand in either region (the partitioned X matrix) As in the single region case, we subtract the partitioned input-output coefficients matrix from the identity matrix I, which is partitioned with zero matrices on the interregional elements, and invert. This gives us the partitioned interregional Leontief inverse L: L PQPR = ( I* 0 0 I E A** A *E A E* A EE ))* (2a) For the general case where there are multiple regions r, s = 1, T, the Leontief inverse is reported as: 53

54 L PQPR = *V l 07 ** l 07 X* l 07 XV l 07 W* WV l 17 l 17 *W l 01 XW l 01 WW l 11 (2b) Each element l XV 07 of the Leontief inverse indicates the output required from sector i in region r to meet one monetary unit worth of sector j final demand in region s. The column totals give us the interregional output multipliers of each sector j. When r = s then the sectors are within the same country and the sum of column entries in this sub-matrix give us own-country output multiplier effects. However, note that even though A XX will be the same as the input-output coefficients matrix of region r XX in an SRIO, this does not mean that L PQPR is necessarily the same as the single region Leontief inverse of region r. This is due to the fact that IRIO also captures interregional feedback effects: that is, intermediate goods produced in region r that are exported to intermediate consumption to another region s before the outputs of region s sectors are imported as inputs by region r sectors. When there are more than two regions, the final demand matrix for total final demand for the output of each sector j in each region s (row totals of vector Y) is the following: DY PQPR = Y V 0 = 0 Y W V 0 0 y y 7 V y 7 W W y 1 (3a) However, it is also possible to focus on any one specific source of final demand for output by the final consumers in one specific region. In that case the final demand matrix is the following: XV Y M 0 V DY M PQPR = WV 0 Y M = XV y 7M XV 0 0 y 1M y 7M WV WV y 1M (3b) 54

55 Each element y XV V 7M of DY M in (3b) represents the final demand for the output of sector j in region r PQPR that is generated by consumer z in region s. Therefore, the matrix as a whole reflects the demand of final consumers z in region s for output from all the sectors in all the regions included in the IRIO. It is also possible to express the final demand diagonal matrix in a way that it shows the total final demand for the output of sector j in region r that is generated by total final consumption in a given region s (e.g. UK final consumption). In that case the final demand diagonal matrix will be the following: V DY PQPR = Y XV 0 = 0 Y WV XV 0 0 y y 7 XV y 7 WV WV y 1 (3c) For the elements of (3c) we have that y 7 XV = [ XV M\* y 7M. By post-multiplying the diagonal matrix of total final demand (3a) by the Leontief inverse the result is the following matrix: L PQPR DY PQPR = l ** * 07 y 7 l *V V 07 y 7 l 07 l 17 l X* * 07 y 7 l W* * 17 y 7 XV y 7 V WV y 7 V l *W W 01 y 1 l XW W 01 y 1 l WW W 11 y 1 (4) Studying the elements of 4 which could also be calculated by using subsets of final demand from 3b and 3c - it is possible to identify how the total production in each sector is ultimately supported or driven by demands for the outputs of different sectors located in different regions. Moreover, 4 allows us to consider these demands in terms of total or any given sub-type of final demand (where subelements of the total y V 7 are applied). Each element l XV 07 y V 7 of (4) represents the production required from sector i in region r to meet the final demand for output of sector j in region s. Examining the elements along each row of 4 - the row totals of which correspond to total output of sector i in region 55

56 r - it is possible to consider output supported at different points of each sector s downstream supply chain. This is the production in each sector required to support the final demand for output produced in others, both within the same country and others. Similarly, the elements down a column of 4 the sum of which is the total output across the global economy ultimately driven by final demand for output in the sector in question - detail a sector s direct plus indirect upstream supply chain requirements, extending beyond the limits of the country where that sector is located. It is important to note that, in constructing this system as a full IRIO, it is necessary that the elements outside the main diagonal, where r s, have been derived from actual data, not estimates 10. The amount of detailed data required to produce IRIO tables is rather large, but in applications like hotspot detection and analysis, the increased accuracy provided by IRIO is of paramount importance Environmental IRIO IRIO can be expanded to report the emissions embodied in transactions between industrial sectors of different regions. The first step is to create an E matrix which includes CO 2 emissions coefficients for industries in all included regions. To do so, it is required to have satellite emissions data reported at sector level, for every sector of every country included in the IRIO. E PQPR = E X 0 = 0 E W X e X 0 0 e W e W e 1 (5) Each element e 0 X represents the CO 2 emissions coefficient (or carbon intensity) of sector i in region r, i.e. the emissions (in physical units million tonnes (Mt) of CO 2 in this chapter) generated by sector i in 10 In MRIO the elements outside the main diagonal, A X* for instance, are estimated by pre-multiplying A ** with a coefficients matrix, the elements of which represent the portion of the monetary flow from region r to region 1 over the total monetary flow to region 1, for each of the industry sectors (see Miller and Blair, 2009, pp 91-93). 56

57 region r per monetary unit worth of output. The emissions coefficients are obtained by dividing the total direct emissions of each sector by the sector s total output. By pre-multiplying E PQPR to the Leontief inverse, each emissions coefficient is matched to the appropriate element of the Leontief inverse. The resulting matrix shall be called Emissions multipliers matrix: Emm PQPR = E PQPR L PQPR = e * ** 0 l 07 e * *V 0 l 07 e X X* 0 l 07 e X XV 0 l 07 e W W* 1 l 17 e W WV 1 l 17 e * *W 0 l 01 e X XW 0 l 01 e W WW 1 l 11 (6) The column totals of (6) for each sector j correspond to the output-emissions multiplier of each sector. However, with the decomposition approach adopted here, (6) allows us to consider the sectoral and spatial composition of these multipliers. Each element e X 0 l XV 07 shows the emissions generated by sector i in region r to meet one monetary unit worth of final demand for the output of sector j in region s. Postmultiplying then with the diagonal (total) final demand matrix (3a), the result is the EIRIO CO 2 emissions matrix Cem PQPR : Cem PQPR = Emm PQPR DY PQPR = e * 0 l ** * 07 y 7 e * 0 l *V V 07 y 7 e * 0 l * 07 y 7 e X 0 l 07 e W 1 l * 17 y 7 e W 1 l 17 XV y 7 V WV y 7 V e * 0 l *W W 01 y 1 e X 0 l XW W 01 y 1 e W 1 l WW W 11 y 1 (7) Cem PQPR is the core matrix of the method used in this chapter and the hot-spot analysis will be conducted on its elements and the version based on (3c) for total UK final demand. The elements of the Cem PQPR demonstrate (for the accounting year in question) the spatial and industrial distribution of emissions embedded in the supply chain of the total domestic final consumption in any one consuming region. Each element e X 0 l XV 07 y V 7 tells us the emissions generated by sector i in region r to meet the total final demand requirements for output of sector j in region s. As with output in equation (4) in the previous sub-section, the elements along each row of (7) show how the generation of emissions in each producing sector i can be distributed among all the sectors j, in all T regions in terms of supporting their final demand, i.e. the downstream supply chain. That is, the elements of (7) report emissions embodied in output to meet final demands of each sector j that are actually generated by sector i. The sum of 57

58 each row in (7) is the total emissions directly generated by each sector i in each region r as would be recorded under a standard PAP measurement. On the other hand, the elements down each column of (7) show the embodied emissions in each point of each sector j s upstream supply chain, regardless of the region where that point is located. Thus, the sum of each column shows the global CO 2 footprint of production to support final consumption (regardless of the location of that final consumption) of each sector j. However, if the focus is to calculate the emissions attributed to a particular country under CAP, it is necessary to: (a) limit the y elements used in calculating (7) to total domestic final consumption generated from within the country in question (but which will involve positive entries in all regions that there are direct imports from); and (b) add the emissions directly generated by those final consumers (generally limited to households with direct emissions generated; in public sector activity recorded in government production rather than consumption activities). Here we focus our attention on the composition of industrial emissions so we limit our attention to (7), whether for total final consumption demands or different types and/or locations of demand therein (i.e. we abstract from emissions directly generated by final consumers). The advantage of the decomposed approach detailed above is that it enables to study the structure of industrial emissions and also the identification of those elements of (7) that make the most significant contribution in terms of CO 2 emitted. Moreover, as shown in the previous chapter, quite often the majority of the CO 2 emissions required by a sector (directly or indirectly) are located within a small number of components of its supply chain. 58

59 3.3.3 Emissions hot-spots This chapter extends the methodology developed in a single region context in the previous chapter to identify hot-spots both upstream and downstream on global supply chains. As in the single region case three types of hot-spots are identified: (a) A sector that in producing output directly generates significantly more emissions compared to other sectors in an economy either to support total final consumption demand or components thereof (e.g. in our hotspot analysis focusing on the global supply chain serving a particular type or location e.g. UK below - of consumer(s)); i.e. has a larger sum of its row in (7). (b) A sector where the output produced to meet final demand for its output (again, either in total or components thereof), directly and/or indirectly, has a larger footprint, i.e. larger sum down its column in (7), compared to other sectors in an economy (c) A point in a sector s downstream or upstream supply chain, an element within (7) that embodies emissions above a set threshold level in serving all or particular type(s) of final consumption demand. Table 1 is a simple illustrative example for two regions, A and B, with 3 industrial sectors in each. Examining the data of Table 1 will aid in understanding the methodology used to identify hot-spots. 59

60 To identify Type (a) hot-spots all that is required is to sum the elements along the row of each sector to calculate the total direct emissions generated by each sector. In Table 1 the sector with the largest volume of direct emissions is Sector 1 of Region B, which can be identified as a Type (a) hot-spot. Similarly, to identify Type (b) hot-spots it is necessary to sum the elements down the column of each sector, calculating the emissions generated throughout the upstream global supply chain to meet each sector s final demand, i.e. the CO 2 footprint of each sector s production to meet final consumption demand. It can be seen in Table 1 that the sector with the largest footprint is Sector 2 of Region B, however Sector 2 of Region A has a similar footprint. Therefore, both sectors can be identified as Type (b) hot-spots. As can be seen in Table 1, but was also suggested by the findings in the previous chapter, Type (c) hotspots more often than not either have the major share of a sector s direct emissions or contribute the majority of emissions to a sector s footprint. As an illustrative example of a hot-spot threshold, we may identify hot-spots of this type by first identifying the row maximums for each row in (7), in the same way that is was calculated in the previous chapter. Then, if we take the average of row maximums, every element of (7) above this average may considered a Type (c) hot-spot. In Table 1 such points are the production of Region A Sectors 1 and 3 and Region B Sector 1 for their own final demand, the production of Region B Sector 1 required by Region A Sector 2 and Region B Sector 2, as well as the output of Sector 2 in Region A required by Sector 2 in Region B. In practice there may be some more specific and policy-motivated means of specifying thresholds (in the context of emissions targets etc.). In the previous chapter it was indicated that hot-spot thresholds can be set in accordance with environmental research outcomes and/or derived from the goals set for each country under international climate change agreements. Unfortunately, not every country faces the same challenges, participates and ratifies the international agreements or has the same agenda in terms 60

61 of the relationship between economic expansion and environmental protection. Under those constraints, the flexibility of this method in determining the hot-spot threshold level is useful. For the purposes of this thesis the threshold is assumed to be the same across every country, however, it is possible to assign a different threshold for each country. In this way the hot-spots identified are examined under the prism of the obligations of the country where they are located, hence facilitating multilateral co-operation. However, the core objective of deriving the method here is to help understand the structure of emissions serving all or particular types/locations of final demand. Focusing on aggregate figures for CAP and/or PAP deprives us from important information on where the majority of emissions to support any given component of county level or total global consumption demand are located and to consider this in the context of understanding domestic and global supply chain relationships. For instance, in Table 1, most of the emissions in the upstream supply chain of Sector 2 in Region 2 are located in the two Type (c) hot-spots. However, to have an even fuller understanding of the emissions it is necessary to apply the hot-spot methodology on the Emissions multipliers matrix (6). Assuming everything else remains constant, identifying hot-spots on (6) - i.e. based on the direct and indirect emissions intensity per average unit of output required at a particular point in an industry s supply chain - enables to locate potential hot-spots in absolute numbers in the event that associated final demand increases (though it is important to note that this involves assuming that average multiplier relationships given by the accounting framework for particular point in time will apply in terms of marginal impacts). Furthermore, studying the underlying multipliers of the hot-spots identified in CO 2 emissions matrix (7), allows for a distinction between those hot-spots that were mainly driven by the multipliers (intensity) and those that the main driver is scale of economic activity. 61

62 3.3.4 Data For this study the IRIO account used is the pilot OECD Inter-Country Input-Output Database focusing on the most recent data of The database consists of: 57 countries, both OECD and non-oecd members, plus the Rest of the World (see Appendix 2.A for a full list of countries); Industrial sectors have been grouped into 37 sectors following ISIC v3.1 (see Appendix 2.B for complete list of sector grouping). Apart from the intermediate goods/inputs the database also includes: Taxes less subsidies on products Cost, insurance and freight price/free on board price adjustments on exports Direct purchases abroad by residents (imports to final consumption) Purchases on the domestic territory by non-residents Value-added at basic prices International transport margins. Final demand is aggregated into five categories: Private (Household) Consumption Non-Profit Institutions Serving Households Government Final Consumption Gross Fixed Capital Formation Inventory (changes in stocks). A key point is that this dataset meets the requirements described by Turner et al (2007) as necessary for a global Inter Regional Input Output (IRIO) table that can be used for multiplier-based CAP and PAP analyses. The database includes direct imports of final goods as well as detailed data on the import of intermediate goods. The data have been harmonised in terms of making consistent data from a range of 62

63 different sources (in particular, building up interregional elements from data on bi-lateral trade flows) and follow the same classification throughout the dataset. The final result is an IRIO table that demonstrates all the transactions between the countries included in IO format. However, since the database is at a pilot stage, it is constantly evolving. This means that there could be inaccuracies, which as the project develops are being reduced in an effort to create a more solid dataset. Nonetheless, the OECD database is preferred in this study over other widely used datasets e.g. WIOD. The most significant advantage of the OECD database is the greater degree of sectoral detail, 37 sectors instead of 35, which according to Wyckoff and Roop (1994) enhances the accuracy of the final results. Moving forward, to create the satellite emissions account it was necessary to explore the emissions directly associated with industrial outputs in the IO table. The account that was built for the purposes of this thesis includes the emissions generated by fuel combustion either during production or by autoproducing heat and electricity, fugitive gases during coal and oil extraction and emissions by industrial processes. Appendix 2.C provides details on how the account was created. The data sources used are IEA fuel combustion data and UNFCCC. The creation of an emissions account was necessary as the number of countries included is larger than any existing dataset and in addition a wider variety of pollution origins has been included to increase accuracy. 3.4 Results General overview Examining the data when we calculate (7) using (3a) for total final consumption demand across all countries reveals some rather interesting findings. Over 85% of the total emissions are located on the main diagonal of sub-matrices of the CO 2 emissions matrix (7), where r = s. This means that 85% of the 63

64 total global emissions are generated by industries producing to meet their own final demand, or in supporting production to meet final demand in industries operating within the same country (although that final demand may in some cases be largely located outside the country). This is true for developed and developing countries alike. In major OECD economies of Germany, UK and USA the respective percentages are 79.4%, 85.9% and 92.4%, whereas in the developing economies of China (excluding Hong Kong which is reported as a separate country) and India the figures stand at 89.5% and 88.8% respectively. If the focus of study is the impact of economic activities within a single country, then IRIO is not necessary as SRIO can provide the necessary information and often with an increased level of detail. The benefit of using an IRIO is that it allows the study of the off-diagonal sub-matrices of the CO 2 emissions matrix (7). It provides us with the opportunity to identify hot-spots located on the international part of any sector s downstream and upstream supply chains, even if the overall impact of these may be small relative to own-country effects on the diagonal of r = s sub-matrices in (7). It also allows us to capture any inter-regional feedback effects, where production sectors in region r export to intermediate sectors in region s with outputs of the latter then imported back to the production sectors in the first region. To demonstrate the ability to study the off-diagonal sub-matrices of (7), here we focus on the UK as a case study. To calculate the CO 2 emissions matrix (7), the diagonal matrix of final demand (3c) was used for s =UK. This means that (3c) shows the output of every sector in every country required by all UK final consumers (i.e. all individual types z=1,, Z across UK households, government, capital formation etc.; or all five groups listed in Section 3.3.4). Therefore, in this case (7) shows the emissions generated globally but ultimately driven by UK total final demand. Data show that UK total domestic final demand was the driver of just over 1,167 Mt of CO 2 in the accounting year of 2009, i.e. the sum of all elements in (7). This equates to a UK carbon footprint in terms of global industrial emissions (i.e. excluding direct emissions by UK consumers) of 1,167 Mt of CO 2, which compares to UK industrial PAP emissions (i.e. the sum of the rows of (7) for r =UK, when calculated using (3a)) of Mt of CO 2. 64

65 The following chart is a rough representation of the interpretation of different elements of (7), under the assumption that the focus is still UK s total final demand (see Appendix 2.A for list of countries where the UK abbreviated by OECD to GBR although representing whole of UK - appears around half way down the list so that we represent in a corresponding position in Chart 1). Of the 1,167 Mt total amount of emissions in 7 for UK final consumption, 714 Mt of CO 2 or 61% was directly generated by UK production sectors, i.e. rows totals of 7 where r =UK, areas 1 and 2 in Chart 1. This includes 1.89 Mt of CO 2 emissions embodied in exported intermediate goods produced by UK sectors, which in terms of elements of (7), these are the elements located on the rows where r =UK but s UK, i.e. the areas labeled 2 on Chart 1. These are emissions generated in the UK to support the production of goods and services in sectors outside the UK that are imported by UK final consumers. Data show then that the majority of direct emissions by UK sectors, Mt of CO 2, were generated to support the final demand of UK sectors, i.e. r = s =UK in 7 or area 1 on Chart 1. Areas 3 and 1 in Chart 1 are where s =UK and represent the footprint of UK sectors serving UK final demand. The footprint of UK sectors is Mt of CO 2. As shown above 79.3% of these emissions are generated by UK sectors, i.e. area 1 in Chart 1 where r = s =UK. The remaining 20.7% of emissions ( Mt of CO 2 ) are generated by non-uk sectors to support the UK total final demand of UK sectors. These sectors are located in areas labelled 3 in Chart 1 and they are the elements of 7 with r UK and s =UK. Finally the sectors in areas 4 of Chart 1 represent emissions by non-uk sectors that produce output to support the UK total final (direct import) demand for output from non-uk sectors, i.e. sectors 65

66 with r, s UK in 7. The total emissions of these sectors are Mt of CO 2, which is a 22.9% share of the total emissions driven globally by UK total final demand. The first points of focus in this section are the sectors outside areas 1 and 2 in Chart 1 i.e. non-uk emissions required by UK final demand. By examining the sum or each sector s row we identify the Type (a) hot-spots located outside the UK. Analysis of results for 7 shows that China s Electricity, Gas and Water Supply (i.e. where r =China and i = Electricity, Gas and Water Supply ) is the largest Type (a) hot-spot outside the UK in terms of emissions driven by UK total final demand. Focusing then on this specific sector we are moving forward by investigating Type (c) hot-spots on China s Electricity, Gas and Water Supply international downstream supply chain, i.e. the elements of (7) located where r =China, i = Electricity, Gas and Water Supply and s China. The reason for this focus is that an analysis of the Chinese Electricity, Gas and Water Supply domestic downstream supply chain, i.e. where s =China, can also be conducted using the SRIO described in the previous chapter. Furthermore, analysing the elements of (7) we can rank the different sectors in different locations in terms of the composition of the footprint of serving UK final consumption demand i.e. the sum of each sector s column in (7) - regardless whether they are located within the UK (s =UK) or outside the UK (s UK). Analysis of the results of 7 shows that the largest Type (b) hot-spots driven by UK total final demand are UK-based sectors. Amongst them, global emissions to support UK final demand for the UK s Health and Social Work is the second largest Type (b) hot-spot behind UK s Electricity, Gas and Water Supply. UK s Health and Social Work is a rather interesting case though, due to the number of Type (c) hot-spots on its international upstream supply chain, i.e. elements of (7) where s =UK and j = Health and Social Work but r UK. Therefore, we focus our investigation for Type (c) hot-spots on UK s Health and Social Work international upstream supply chain. 66

67 As shown above, emissions generated by UK production sectors are the major contributors to UK sectors footprint driven by UK total final demand. However, there are UK sectors where each monetary unit worth of final demand has a larger impact on the non-uk side of their upstream supply chain. That means that the sum of the elements down the sector s column in 7 and the underlying emissions multiplier matrix 6 are larger on the non-uk rows rather than the rows of UK sectors. Such examples are UK s Motor Vehicles, Trailers and Semi-trailers and Office, Accounting and Computing Machinery. In general, it is important to note (particularly in terms of useful policy analysis tools that could be extracted from the IRIO framework) the total footprint of serving UK final consumption demand for each sector j in each region s could also be calculated by multiplying that sector s Type I emissions multiplier (column total from equation 6 ) with the sector s total UK final demand. Sub-totals for elements of the multiplier located in different countries could be used similarly. This builds on the familiar use of multiplier values to assess particular types of impact in particular areas whenever there is a change in economic activity. However, using the adopted methodology of this chapter, post-multiplying (3c) to (6), enables us to study and analyse the structure of the footprint in detail. In practice what this approach essentially involves is multiplying the total final demand in question (with our focus here on the total of UK final demand across all five types identified in Section 3.3.4) for each production sector in each region with every element down the sector s emissions multiplier column in (6). However, we do present examples of results, for example in Table 2 below, where users of the research output can conduct simpler multiplier calculations. 67

68 3.4.2 Type (a) and downstream Type (c) hot-spots outside the UK driven by UK total final demand As already discussed in the previous sub-section, the first focus point of this chapter is to locate Type (a) hot-spots outside the UK. In total the non-uk sectors generate Mt of CO 2 (row totals of 7 excluding r =UK) Table 2 shows the Top 10 sectors in terms of direct emissions associated with UK total final demand that are located outside the UK. In Table 2 the Direct Emissions column (first column of results) refers to the sum of the elements along each sector s row in (7), while the next column shows these emissions as a percentage share of the total emissions generated driven by UK total final demand. The third column indicates the share of the sector s total direct emissions (full PAP) that are the UK-driven entries in the first column. The CO2 Intensity and UK Total Final Demand columns refer respectively to the sector s elements on E matrix (5) - i.e. the direct CO 2 intensity e X 0 of each sector - and on final demand matrix (3c) respectively, i.e. the final demand from UK element for y XV 7 of each sector. Please note that the CO 2 intensity is in Mt/$m of output. The unit used might make the figures of that column seem rather small, however, they represent significant volumes of emissions that should not be neglected. The final column refers to the monetary value of the output of each sector that is ultimately supported/driven by UK total final demand. This is the sum of each sector s row in (4) when calculated using (3c). If we multiply this against the direct CO 2 intensity of the sector, we have another means of generating the result in the first column (but one that is embedded in calculation of (7), that is considering the supported output multiplier effect rather than moving straight to the emissions multiplier). The difference between the figures of columns 5 and 6 in Table 2 indicates whether each sector mainly produces final goods for UK final consumers or intermediate goods to support other sectors UK total final demand. 68

69 As reported in the discussion in the previous section (and also reported at the bottom of Table 2) the total direct (PAP) emissions generated globally driven by UK s total final demand are 1,167.2 Mt of CO 2. Of these emissions Mt of CO 2 are generated outside the UK, i.e. 38.8% of the total direct emissions driven by UK total final demand. The sectors listed in Table 2 account for 37% of the emissions driven by UK total final demand and generated outside the UK. The vast majority of the sectors listed on Table 2 most notably Electricity, Gas and Water Supply in different countries - have minimal amounts of UK total final demand compared to their output associated with UK total final demand. This implies that the output of the i = Electricity, Gas and Water Supply (hereafter EGWS) sectors in the countries shown on Table 2 is used as input by other sectors in these countries (assuming a low level of trade in EGWS itself, though gas exports may be important) that either export final goods to the UK, or produce outputs to intermediate demands entering supply chains that ultimately (but indirectly) serve UK final demand. That is, there may be many rounds of multiplier effects involved. This is what our Type (c) hotspots allow us to consider. 69

70 70

71 Before we turn our attention to downstream Type (c) hot-spots, let us consider the importance of direct CO 2 intensities. Of the ten sectors in Table 2 the ones that directly generate the most significant amount of CO 2 emissions, i.e. sum of sector s row in CO 2 emissions matrix (7), are the Chinese EGWS, the USA s Transport and Storage, the Chinese Basic Metals and the Russian EGWS. They have the most significant shares of the total direct emissions driven by UK total final demand and can therefore be considered as Type (a) hot-spots in the global supply chain serving UK consumption. The largest Type (a) hot-spot of Table 2 is China s EGWS sector. Figures in Table 2 show that the main driver of the emissions generated by the sector is the CO 2 emissions intensity. China s EGWS CO 2 emissions intensity is the 24th highest amongst all 2146 sectors (with 37 industries in 58 regions/countries including ROW) included in the OECD ICIO framework used. Even though the sector s output associated with UK total final demand (the figure on the Total Output for UK Final Demand column on Table 2) is ranked only 100th amongst all 2146 sectors included in the framework, still, due to the relatively high emissions intensity, the direct emissions of China s EGWS driven by UK total final demand are the largest outside the UK. In fact, the Chinese EGWS sector is ranked 4th in direct emissions driven by UK total final demand amongst all 2146 sectors. However, the results reported in Table 2 suggest that it is rather common for EGWS sectors to be relatively CO 2 intensive. In fact, the only exception is the German EGWS sector, which in 2009 had a mixture of production technologies that allowed for a rather low CO 2 emissions intensity, lower than the other relevant sectors of Table 2. The reason why Chinese EGWS tops Table 2 is that at the same time it has the second largest direct emissions intensity and the largest output associated with UK total final demand amongst all the EGWS sector of Table 2. Focusing on the top direct emitter of Table 2, Chinese EGWS, the significant difference between the sector s UK total final demand and the output produced due to UK total final demand indicates that the majority of the sector s direct emissions is distributed along the sector s downstream supply chain. 71

72 Quite possibly then, there could be important Type (c) hot-spots on China s EGWS downstream supply chain. Here, for an element of the CO 2 emissions matrix (7) to be considered as a Type (c) hot-spot, when UK total final demand is used, we take a simple threshold level as the average of row maximums, which works out at 0.29 Mt of CO 2. Table 3 shows the Type (c) hot-spots on China s Electricity, Gas and Water Supply downstream supply chain to support UK final consumption that are located outside the Chinese borders, i.e. r =China, i = Electricity, Gas and Water Supply and s China. These hot-spots are elements of (7) that were summed to calculate the China s EGWS direct emissions presented in Table 2. Apart from excluding the hot-spots located in the domestic downstream supply chain of the Chinese EGWS, there are no other restrictions as to where the hot-spots might be located. Therefore, the Type (c) hot-spots presented in Table 3 are in fact the only ones in the international part of China s EGWS downstream supply chain. The first results column in Table 3 shows the element of the respective sector on the row of Chinese EGWS in (7), which corresponds to a specific point in Chinese EGWS downstream supply chain, while the second column shows these elements as a percentage share of the total direct emissions of China s Electricity, Gas and Water Supply. The third column is the element of each sector listed in Table 3 on 72

73 the row of Chinese EGWS in (6) whereas the fourth column shows the y 7 XV in final demand matrix (3c) for each of the sectors listed in Table 3, i.e. the UK final demand for each of the sectors in Table 3. Interestingly enough, all the Type (c) hot-spots on the Chinese EGWS sector row of (7) that are associated with UK total final demand and located outside China, are found within the UK. In total they have just over a 22% share of the total Chinese EGWS emissions that are attributable to UK final consumption. It can be seen that the top 4 Type (c) hot-spots of Table 3 have a more significant share (14.16%) of Chinese EGWS direct emissions, compared to the other Type (c) hot-spots of Table 3. Examining the figures of Table 3 reveals that the emissions embodied in the top 4 Type (c) hot-spots are driven by the volume of consuming sector s total UK final demand rather than their emissions multipliers, which are well below the emissions multipliers of other sectors in Table 3. This is not surprising given that UK total final demand is mainly served by UK sectors (i.e. 87.8% of UK total final demand is expenditure in UK sectors). In fact, the top 4 sectors of Table 3, UK s Health and Social Work, Construction, Public Admin and Defence; Compulsory Social Security and Wholesale and Retail Trade; Repairs ; are also within the top 5 sectors in terms of total UK final demand, the other one being UK s Real Estate Activities Type (b) and upstream Type (c) hot-spots driven by UK total final demand As seen in a previous sub-section, UK total final demand is primarily met by the output of UK sectors. This being the case, one could argue that when looking for the sectors with the largest CO 2 footprint driven by UK total final demand, the majority of them will also be UK-based. Table 4 shows the top 10 sectors in terms of footprint, i.e. sum of each sector s column in (7), driven by UK total final demand. 73

74 74

75 In Table 4 the first column is the sum of the elements down each sector s column in (7). The second column is the share of each sector s footprint of the total global emissions driven by UK total final demand. The third column shows the footprint of each sector driven by UK total final demand as a percentage share of the sector s footprint driven by global total final demand. Column 4 is the sum of XV the emissions multiplier elements down each sector s column in (6) and finally column 5 shows the y 7 in (3c) for each of the listed sectors, i.e. each sector s UK final demand. As expected, the top 10 sectors with the largest footprint driven by UK total final demand are all UK based. The non-uk sector with the largest footprint driven by UK total final demand is the Chinese Textile, Textile Products, Leather and Footwear, which is ranked 12th amongst all the sectors in terms of footprint driven by UK total final demand and thus not included in Table 4. Examining the sectors of Table 4 there is a common trend across the majority of them. The footprint of these sectors, driven by UK total final demand, has over a 90% share of the sectors footprint driven by global total final demand (i.e. the column total of (7) calculated using (3c) as a share of that calculated using (3a)). Given that the Type I emissions multiplier is constant regardless of the location of the final consumer, these figures show that the final demand requirement of the sectors in Table 4 largely originates within the UK itself. Of all the sectors listed in Table 4, UK s Electricity, Gas and Water Supply has by far the largest footprint in serving UK final consumption (and generally if we use (3a) to calculate (7)). This is mainly driven by the sector s Type I emissions multiplier (i.e. the sum of the sector s column in (6)) which is the largest amongst the sectors of Table 4. On the other hand, it can be seen that there are UK sectors like Health and Social Work and Wholesale and Retail Trade; Repairs where the magnitude of the footprint is driven by the volume of their total UK final demand rather than the (direct plus indirect) CO 2 intensity given by the emissions multiplier. More generally, for the majority of the sectors on Table 4 the main driving factor is indeed the value of their total UK final demand rather than their Type I emissions multiplier. The policy implications of this information are that for the majority of the sectors in Table 4 it 75

76 would be preferable to explore environmental policies that are associated with consumer behaviour instead of trying to de-carbonise their upstream supply chains. For example, educating the general population in making more efficient use of the services of the Wholesale and Retail Trade; Repairs sector could lead in reduction of the sector s final demand. The emissions related consequence of this reduction would be reduced embodied emissions in the sector s upstream supply chain. However, as discussed in the conclusion of the previous chapter, the findings of this hot-spot analysis do not provide an overview of all the potential impacts that would come as a result of policies introduced in the sectors of Table 4 (or any other sectors). Further analysis would be necessary to pick those sectors that any decrease in final demand, in order reduce their footprint, would have the least impact possible in value-added lost and increased unemployment. Just as we considered Type (c) downstream hot-spots linked to Type (a) PAP hot-spots in the previous section, it is worth investigating the upstream supply chains of the top sectors of Table 4 to see whether there are any interesting and/or important Type (c) hot-spots. This involves considering column entries of (6) and (7) for the sectors identified in Table 4. First, for the UK EGWS sector the major contributor to the sector s footprint is its own-sector emissions to meet its own total UK final demand, which embodies 219 Mt of CO 2, i.e. almost all of the emissions of the Type (b) hot-spot. However, it is worth noting the level of aggregation involved in the OECD EGWS sector. Water supply tends to be electricity intensive while electricity production can be gas-intensive. Therefore, there are likely to be important inter-sectoral effects hidden in the own-sector (i = j, r = s) EGWS results throughout our results for both the multiplier effects in (6) and total supported emissions in (7). This is an issue that we return to in the next chapter, where sub-national regional data are used to further analyse hot-spots identified via IRIO analyses. In general, though, a single region analysis based on more sectorally disaggregated published regional or national accounts would tend to separately identify what tend to be relatively energy- and emissions-intensive utilities sectors. 76

77 On the other hand, the second largest sector of Table 4, UK s Health and Social Work, has a more interesting upstream supply chain as it includes several Type (c) hot-spots located outside the UK territory. Table 5 shows the Type (c) hot-spots in UK s Health and Social Work that are driven by UK total final demand and located outside the UK. These are elements of (7) with s =UK, j = Health and Social Work and r UK. As a reminder our illustrative threshold level for a Type (c) hot-spot is 0.29 Mt of CO 2 and all the entries in Table 5 are above this level. The first column of Table 5 includes the element of each sector in Table 5 on the CO 2 emissions matrix (7) - i.e. column entries for j = Health and Social Work and s =UK when (7) is calculated using (3c) - while column 2 presents them as a percentage share of UK s Health and Social Work footprint in serving UK final consumption demand. Column 3 includes each sector s element on the Leontief inverse 77

78 (2b) while column 4 shows the e X 0 of E matrix (5) for each of the producing sectors in Table 5. The elements of (2b) and (5) are presented separately and not as elements of (6). The benefit of using this approach is that we can distinguish whether the receiving sector, in this case UK s Health and Social Work, requires large volume of output from any one producing sector or whether it is the producing sector s emissions intensity that drives the emissions of that Type (c) hot-spot. Please note that Table 5 presents the Type (c) hot-spots in a different way compared to Table 3. The reason is to demonstrate the different analysis options when using an IRIO. As can be seen from Table 5, a rather large number of Type (c) hot-spots can be found on UK s Health and Social Work upstream supply chain that are located outside the UK and driven by UK total final demand. Their total contribution to the sector s footprint is just over 21% 11. Analysing the UK s Health and Social Work Type I emissions multiplier it can be seen that each monetary unit of final demand has a more significant impact within the UK. 61% of the emissions embodied in the sector s footprint are generated by UK based industries and 39% abroad. The results in Table 5 encompass most of this 39%. The results in Table 5 imply that the UK s Health and Social Work has some rather specific upstream international supply chain requirements, which will involve both direct imports and multiplier impacts of other intermediate input (domestic and imported) requirements. Although located in several different trading partners of the UK, the CO 2 -emitting outputs required come from Chemicals and Chemical Products, Coke, Refined Petroleum Products and Nuclear Fuel, Electricity, Gas and Water Supply, Transport and Storage and Basic Metals sectors in various countries around the world. The appearance of these sectors in our hot-spot analysis may be expected given that their activities include 11 For most of UK sectors the majority of emissions generated to support their final demand are located within the UK. However there are three sectors, Motor Vehicles, Trailers and Semi-trailers, Office, Accounting and Computing Machinery and Machinery and Equipment n.e.c. that the main body of the emissions generated to support their final demand is located outside the UK. For each of the aforementioned UK sectors the contribution to the Type I emissions multiplier from abroad is 60%, 57% and 53%. Still due to the relatively small volume of total UK final demand their footprint is rather small compared to other sectors and thus not featured in Table 5. However, assuming that everything else remains constant, an increase in the total UK final demand of UK s sectors C34, C30 and C29 would lead to a significant increase in the size of their footprint, the majority share of which would be outside UK borders. 78

79 the production of pharmaceuticals, diesel, gas and precious metals as well as their transportation. These are all products that are necessary for Health and Social Work activity. However, they may not be the obvious focus of attention in considering how to address the carbon footprint of this type of sector. Table 5 suggests that Health and Social Work (hereafter HSW) mainly depends (directly or indirectly) on production of output in the global Chemicals and Chemical Products industries (hereafter CCP). HSW sectors of the different countries in the OECD database tend to have highest output multiplier values located in CPP sectors i.e. elements for i =CCP and j =HSW in the inter-regional Leontief inverse in equation (2). At the same time the direct CO 2 emissions intensity of CCP does not vary greatly from country to country. Therefore, the differences in the embodied emissions associated with CCP production in the respective Type (c) hot-spots are largely associated with the output multiplier relationship with UK HSW. One exception is the requirements from German CCP. As can be seen in the third column of Table 5 the output multiplier of German CCP is larger than the output multiplier of the CCP sector in the USA. This implies that UK HSW requires larger volumes of German CCP output to support its domestic final demand. However, the USA CCP CO 2 intensity is twice as large as the one of the German CCP (data in column 4 of Table 5). As a result, the Type (c) hot-spot where the producing sector is USA CCP has more embodied emissions than the one where the producing sector is German CCP. On the other hand, there can be seen significant variations in underlying determinants that are not limited to the output multiplier effect when it comes to EGWS hotspots in the UK HSW supply chain. For instance, the third column of results in Table 5 shows that the Chinese EGWS hot-spot (the largest in the table) is driven largely by this sector being more (directly) emissions intensive than any other sector in Table 5, rather than the level of output requirements. This is further illustrated by the fact that even though the UK HSW sector has somewhat similar output requirements for EGWS from China and The Netherlands ( $m per unit of output to meet final demand relative to in the third 79

80 column), still the difference in direct emissions intensity ( Mt per $1m output relative to ) puts the Chinese Type (c) EGWS hot-spot at the top of Table 5 whereas the Dutch one is second to last. From a policy perspective the knowledge of the structure of embodied emissions of any given sector could provide policy makers with important information to inform additional options for targeted policies in reducing the carbon footprint of that sector. However, in the case of the Type (c) hot-spots of Table 5 there could be jurisdiction issues due to the fact that industries of different countries are involved. Still the knowledge acquired from hot-spot analysis on an IRIO level can be used on a commercial level. For example, firms that operate within UK s Health and Social Work could apply commercial pressure to their suppliers abroad, in an effort to reduce their CO 2 footprint. This information may also be of use to procurement managers in public run Health and Social Work activities. It is quite often the case that purchase decisions will focus on the economic side of the purchases, looking for those imports that meet the needs and requirements at the minimum cost. However, where there is a real need and commitment to reduce the carbon footprint of public sector activities (which generally focusses on more direct sources of emissions, such as energy efficiency of buildings) having access to the type of information reported in Table 5 could help add the element of environmental impact in the decision process. 3.5 Conclusion and extension The use of an IRIO enables a more accurate calculation of the emissions attributed to each sector especially under a Consumption Accounting Principle (CAP). In a SRIO, if we were to estimate the emissions embodied in the imports of any sector, it was necessary to make some generalising assumption, such as that all the trade partners of the country under examination were using the same production technology at the same point in time. As more countries are included in the IO framework, 80

81 we obtain more detailed data on the environmental impact of the sectors within these countries. Therefore, the number of countries for which we need to assume that they share production technologies is gradually reduced and the results we obtain better reflect the embodied emissions in any sector s upstream supply chain. Furthermore, in IRIO imports and exports of intermediate goods are endogenous, rather than exogenous inputs and exported final demand, and as a result the multiplier effects can be calculated more accurately. Expanding the hot-spot methodology to a global IRIO framework enables the identification of hotspots beyond the borders of a single country. It is possible to highlight components of the international side of the downstream and upstream supply chain of any sector and study the impact that final demand of any sector has outside the borders of the country where the sector (and/or final consumption demand for output) is based. However, the findings of hot-spot analysis on an IRIO need to be reviewed with some degree of attention. Any kind of IO analysis is heavily dependent on the quality of the data used. This is even more important in IRIO, where the data come from various different sources, with different collection procedures and techniques, a point that was raised by the UK Department of Energy and Climate Change (DECC) (as reflected in the report by the House of Commons Energy and Climate Change Committee, 2012a). As such it is impossible to be absolutely sure that the quality of the data used to compile the IRIO tables is the same across the board. For the purposes of this thesis the OECD Inter-Country Input Output database (OECD, 2015) was used. The creators of the database in OECD had to reconcile and balance the data from all the different sources in order to create a credible dataset. However, it is rather common in large IRIO datasets like WIOD and the OECD Inter-Country Input Output database that the industrial sectors are highly aggregated in order to achieve a uniform classification across all regions. Over-aggregation can lead to analytical errors (which will be discussed in the next chapter) while at the same time masks the true nature of the Type (c) hot-spots when these involve the production of a highly aggregated sector. For 81

82 example, as it has been mentioned for EGWS throughout this thesis so far, it is impossible to judge which one(s) out the sectors that are aggregated into EGWS contributes the main share of embodied emissions in any of the EGWS Type (c) hot-spots in Table 5 (section 3.4.3). Due to this limitation, hotspot analysis is mostly useful in providing spatial information on hot-spots, which then would need to be further investigated using national and sub-national IO tables in order to get the maximum level of details possible. Still, the development of this type of datasets could gradually lead to the resolution of the data issue that DECC is highlighting. The other point of required attention is the adjustment of the hot-spot threshold level. In this thesis, the average of the row maximums in the CO 2 emissions matrix is used for illustrative purposes to aid in demonstrating how the proposed methodology can be used. This is by no means the optimal way of setting the threshold. As mentioned in the previous chapter it is possible to adjust the threshold either based on environmental research or based on the emissions goals set by international agreements. Given that different countries have different agendas and interests, the latter approach seems more plausible given that participation in and ratification of an international agreement implies that the parties involved accept the goals set and the accounting methods proposed. Generally, the way IRIO hot-spot analysis can influence policies is significantly different to the analysis on SRIO. Whereas at the national level (studied using SRIO) it is possible to regulate any sector s upstream and downstream supply chains in order to reduce their footprint or their direct emissions, at the international level there are jurisdictional barriers and as such bilateral co-operation is necessary. However, having the information from IRIO hot-spot analysis available can lead to indirect measures involving consideration of environmental parameters when purchasing necessary inputs for public sector activities, as discussed for the UK s Health and Social Work sector. Additionally, the same results can be used as a basis for developed countries to provide funding to carbon reduction initiatives in developing countries, under the carbon finance concept. For example, in Table 2 a number of non-uk 82

83 sectors are presented that generate significant emissions due to UK final demand. This information coupled with a carbon price could be considered as UK s mandatory contribution to carbon reduction funds, which in turn will be used by the countries influenced in order to develop carbon saving innovations. From a different perspective, this information enables private firms to become significant contributors in the reduction of their footprint by identifying the most polluting components of their upstream supply chains and therefore acting to enforce the use of more environmentally friendly technologies by their suppliers. It is clear then that performing a hot-spot analysis at the inter-regional level helps with generating additional information that could not be obtained by just focusing on the single region level. Unfortunately, as discussed above, there are specific limitations that derive from the current characteristics of the available IRIO tables. For example, the level of aggregation poses significant limitations in our understanding, especially in the case of EGWS sector which so far has been flagged multiple times as significantly polluting, but for which we cannot be sure which of the different components of this aggregated sector actually holds the largest share of emission or whether the share is evenly distributed. It is important then to identify how significant these limitations are and more importantly how much more information we could obtain by overcoming them. A logical next step then is to apply the proposed methodology on detailed and disaggregated subnational IO tables, published directly from the local authorities rather than derived from IRIO tables (as was the case in Chapter 2). This exercise will help understand the level of details that can be obtained by conducting a hot-spot analysis and at the same time how restrictive and problematic is (or is not) the use of datasets with highly aggregated sectors. Furthermore, should the results from the application on disaggregated sub-national IO tables prove that there are significant errors associated with the aggregation then this will provide a strong argument in favour of the development of disaggregated national and sub-national IO tables. At the moment, as Turner (2006) points out, there are doubts on 83

84 whether the investment on detailed IO datasets is worthwhile in terms of the resources required. Even in cases like Scotland where the detailed IO tables have been developed (and will be used in the next chapter), there has been limited use of those IO tables for emissions related analyses. Applying hotspot analysis on disaggregated IO datasets could then act as reassurance that there are significant gains to be made by using these datasets and as a result encourage more extensive use and continued support/further development of these IO tables. 84

85 Chapter 4: The benefits of using regional Input Output tables and the importance of region-specific satellite emissions data. Abstract Environmental Multi- and Inter-Regional Input Output methods are commonly used tools to account for CO 2 emissions generated by production of and demand for output of economic sectors in different countries and, crucially, to capture the impact of international trade. However, the need for harmonised data can result in a high level of aggregation, which reduces the detail on activity within a given economy and leads to biases and errors. Single Region Input Output analysis based on national/regional statistics can be significantly more detailed in terms of interactions between producing and consuming sectors within the region. The additional details on the regional level can be a useful tool for regional governments that aim to reduce the emissions generated within their borders. However, regional IO tables lose in terms of details and accuracy when it comes to interaction between producing and consuming sectors located in different regions/countries. This chapter argues that Single Region Input Output can be used as a follow on to the type of hot-spot analysis reported in the previous chapter to provide the details lost through the greater level of sectoral aggregation imposed in an Inter-Regional Input Output framework. Moreover, taking the UK as a case study, we consider the implications of moving from an IRIO down to a SRIO analysis for a region where key polluting sectors have distinct pollution characteristics (e.g. Scottish electricity supply). Analysis of Scottish data reveals that the major Scottish direct emitting sectors form important but distinct parts of the aggregated top direct emitting UK sectors as identified using IRIO. However, we find that the emissions hot-spots calculated for some of the Scottish sectors will be overestimated if region-specific emissions data for Scotland are not considered. We demonstrate how region-specific satellite emissions data are crucial to perform 85

86 accurate calculations and that adjusting the emissions intensity of one sector using region specific data has an impact on the CO 2 footprint of all the sectors in an economy. 4.1 Introduction Climate change is a policy problem that has attracted significant research attention from a range of disciplines over the past decades. As explained in the previous two chapter, there have been numerous studies, many involving application of input-output (IO) methods on the impact of human economic activities, especially in terms of emissions of greenhouse gases. On the policy front there have been efforts to mitigate the climate change especially via the multi-lateral decisions made in the conferences of UNFCCC parties. This chapter moves on to consider the context of where agreements made at national level (e.g. UK) involve devolution of responsibility in setting and meeting targets to regions (e.g. Scotland), where the composition and nature of polluting activity may be quite distinct. To recap, IO analysis has become a widely used framework due to the fact that it is a rigorous and systematic approach to study the direct emissions and the footprint, the emissions attributed to final demand, of all the sectors in an economy. Due to the importance of international trade, more recent studies have tended to favour IO frameworks that include multiple regions and capture the environmental impact of international trade, i.e. Multi-Regional (MRIO) or Inter-Regional (IRIO) Input- Output. 12 However due to the nature of the datasets required for IRIO, there are losses in terms of accuracy as sectors need to be aggregated permitting harmonisation across a single dataset incorporating many countries. 12 The distinction between MRIO and IRIO is explained in the previous chapter. From this point onwards the chapter will focus mainly on IRIO. 86

87 In general, there is a trade-off between using Single Region Input Output (SRIO) created from published IO tables based on national/regional statistics and IRIO, which involves harmonising data from numerous sources. In particular, SRIO offers a high level of detail due to the fact that the industrial sectors included in national/regional statistics are generally less aggregated than in IRIO tables. On the other hand, by reporting trade flows in IO format, IRIO allows us to consider the environmental impacts of emissions/energy/water uses embodied in international trade. Focusing on the UK as a country and Scotland as a (devolved) region of the UK (with its own climate change targets 13 ), this chapter argues that there are significant gains to be made by using the results of IRIO as a guideline to identify sectors that would benefit from more in-depth analysis at national or sub-national level using SRIO. Furthermore, it is suggested in this chapter that having region-specific emissions data for key sectors (e.g. electricity supply) operating at regional level within the UK (e.g. Scotland, where generation from renewables is more prevalent than at national level) is of paramount importance. The structure of this chapter is as follows. The second section offers a brief review of relevant literature. The third section briefly presents the methodology used while the fourth section compares the results obtained by using IRIO and SRIO. The fifth section compares the results from SRIO to the emissions reported by the Scottish industries and the sixth section concludes and suggests potential extensions. 4.2 Literature background The extent of international trade in modern economy has led numerous researchers to study its environmental impact. To that end MRIO has been a rather useful and evolving tool. Wiedmann et al (2011) provide an extensive review of the evolution of MRIO. The problem is that to create MRIO and IRIO tables it is necessary to have a common sector mapping as pointed out by Turner et al (2007). This 13 See 87

88 leads to the use of aggregated sectors. Any level of sector aggregation, including over-aggregation, has the potential to generate bias and lead to aggregation errors. Ara (1959) and Miller and Blair (1981), both present criteria that determine whether an aggregation is acceptable or not. Following these criteria, for two or more sectors to be aggregated in an acceptable way, they need to have similar structures in terms of required inputs from other sectors. Even when that is the case though it has been shown by de Mesnard and Dietzenbacher (1995), that the input coefficients of an aggregated sector do not match the weighted input coefficients of the sectors included in the aggregated sector. This finding indicates the potential for errors that is driven by overaggregating sectors. As pointed out by Lahr and Stevens (2002) both the aggregation error as a concept and the methods in which the aggregation bias can be measured have been explored by various studies. Unfortunately though, the method for estimating the aggregation bias presented by Miller and Blair (2009, pp ) requires the availability of both the aggregated and disaggregated IO tables in order to estimate the bias. This is a problem faced by studies where the IO tables have been obtained by external sources that have already aggregated the sectors. Therefore, the underlying assumption when using highly aggregated IO tables is that the people that have created the tables have aggregated sectors in such a way that follows the criteria of Ara (1959) and Miller and Blair (1981). So far the discussion involved aggregating sectors at the economic level. When considering the environmental impact of different sectors, the aggregation criteria can be expanded to include similar emissions. However, according to Hawdon and Pearson (1995), over-aggregated sectors can include industries with significantly different pollution characteristics. As a result, analysis of the data from IRIO tables where, for example, we have a single Electricity, Gas and Water Supply sector in each country, could mask significant levels of emissions involved in key inter-sector interactions, such as the emissions generated producing electricity used to support water supply. Furthermore, different industries with different emissions intensities that are aggregated together in an IRIO may be viewed as sharing the 88

89 same (average) emissions intensity, which would generate misleading results particularly where marginal changes in activity impacting just one sub-industry. Especially in the case of electricity generation several studies argue in favour of a high degree of disaggregation (e.g. McNicoll and Blackmore, 1993; Gale, 1995). In fact, Wyckoff and Roop (1994) have shown that by using more aggregated sectors leads to underestimations in the total CO 2 emissions. Depending on several parameters, including but not limited to funding, national/regional IO tables can less aggregated, although as discussed by Lahr and Stevens (2002) regional IO tables that are not implemented properly can be highly aggregated compared to national IO tables and therefore lead to significant errors. Comparing the OECD Inter-Country Input Output database (OECD, 2015), that was used in the first two chapters (and will also be used in this one), with the database used by Munksgaard and Pedersen (2001), for example, it is easily noticeable that the latter includes far less aggregated sector mapping, 130 sectors (for the tables after 1988) compared to 37 in the OECD database. Therefore, SRIO based on national/regional IO tables with a lower level of aggregation offers the opportunity for analysis without errors due to over-aggregation issues highlighted by Miller and Blair (1981 and 2009), de Mesnard and Dietzenbacher (1995) and Hawdon and Pearson (1995). Unfortunately choosing between the single region framework (SRIO) and the multiple regions (IRIO) includes significant trade-offs. Focusing on one country potentially compromises the accuracy of any findings regarding impacts of international trade, as shown by Lenzen et al (2004) and Shui and Harriss (2006). At the same time the aggregation of every extra-regional region or country into a limited number of aggregated regions, leads to the introduction of bias (Miller and Blair, 1981). It can be seen then that moving from a highly aggregated IRIO to a disaggregated SRIO involves reducing the biases due to the aggregation level of industrial sectors but increasing the bias due to the aggregation of countries and regions. Ideally, when there is an option to do so, one should select an IO framework based on the minimum bias due to aggregation. However, more often than not the selection is directed 89

90 by the research interests as pointed out by Lahr and Stevens (2002). In the case of our hot-spot analysis, focusing on the national/regional level not only benefits from enhanced level of intra-country detail, but the findings may be more relevant to policy makers as they usually focus on territorial production-driven emissions, partly due to the type of jurisdictional issues raised by Turner et al (2011a). 4.3 Methodology Environmental Input Output Recapping on the methodology detailed in the previous chapter, our first step is to use the IRIO method presented by Miller and Blair (2009), Turner et al. (2007). The key transmission mechanism in a demanddriven IRIO, and indeed every IO, model is the Leontief Inverse matrix. For an IRIO it is the following: L PQPR = *V l 07 ** l 07 X* l 07 XV l 07 W* WV l 17 l 17 *W l 01 XW l 01 WW l 11 (1) The Leontief Inverse is the matrix of output multipliers. Each element l XV 07 of the Leontief inverse indicates the output required from sector i in region r to meet one monetary unit worth of sector j final demand in region s. The sum of the elements down the column of each sector j is the Type I multiplier which reflects the value of the output required by all the sectors in all regions to support one monetary unit worth of final demand for the output of sector j in region s. When the Leontief Inverse is post-multiplied with the diagonal matrix of final demand then the result is the following matrix: 90

91 X = L PQPR DY PQPR = l ** * 07 y 7 l *V V 07 y 7 l 07 l 17 l X* * 07 y 7 l W* * 17 y 7 XV y 7 V WV y 7 V l *W W 01 y 1 l XW W 01 y 1 l WW W 11 y 1 (2) Matrix X is the N N output matrix. Each element l XV 07 y V 7 represents the output of sector i in region r required to meet the final demand of sector j in region s; i, j = 1,, N and r, s = 1, T. When the focus is on a single region then r = s. For the purposes of this chapter as final demand is considered total final demand. The IRIO can be extended to report the emissions generated during the production of output. Since environmental impact is considered an externality is usually omitted from national statistics and needs to be calculated indirectly (Leontief, 1970). To achieve this, it is necessary to use emissions coefficients that will transform the standard IRIO so that it reports output in terms of emissions rather than monetary value. The emissions coefficients are arranged in a matrix format, matrix E. E PQPR = E X 0 = 0 E W X e X 0 0 e W e W e 1 (3) Each element e X 0 of matrix E is the emissions coefficient of sector i and shows the emissions generated by each unit of output produced by sector i in region r. To calculate the emissions coefficient of any sector, the total emissions generated are divided by the total output of the sector. For this process it is necessary to have a satellite emissions data at the sector level for every sector in every country included in the IRIO. Pre-multiplying the E matrix (3) to the Leontief Inverse (1) generates what in the previous two chapters was described as Emissions multipliers matrix. 91

92 Emm PQPR = E PQPR L PQPR = e * ** 0 l 07 e * *V 0 l 07 e X X* 0 l 07 e X XV 0 l 07 e W W* 1 l 17 e W WV 1 l 17 e * *W 0 l 01 e X XW 0 l 01 e W WW 1 l 11 (4) Each element e X 0 l XV 07 shows the emissions generated by sector i in region r to meet one monetary unit worth of final demand for the output of sector j in region s. Similarly to the Leontief Inverse, summing the elements down the column of each sector j gives us the Type I emissions multiplier for that sector. This corresponds to the emissions generated by all sectors in all regions in producing output to support the final demand of sector j in region s. Post-multiplying the Emm PQPR by the diagonal matrix of final demand results in the core matrix of the Environmental IRIO (EIRIO), which throughout the previous two chapters was called CO 2 emissions matrix (Cem). Cem PQPR = EL PQPR DY PQPR = e * 0 l ** * 07 y 7 e * 0 l *V V 07 y 7 e * 0 l * 07 y 7 e X 0 l 07 e W 1 l * 17 y 7 e W 1 l 17 XV y 7 V WV y 7 V e * 0 l *W W 01 y 1 e X 0 l XW W 01 y 1 e W 1 l WW W 11 y 1 (5) Cem is the main matrix that will be used for analysis in this chapter. Each element e X 0 l XV 07 y V 7 of (5) shows the emissions generated by sector i in region r to support the final demand of sector j in region s. As with the case of output matrix (2) when focusing on a single region we have that r = s. The elements on the row of each sector i represent the different components of the sector s downstream supply chain, the output produced to support the final demand of other sectors, and the sum of all the elements along the row is the amount of the direct emissions generated by sector i. On the other hand, the elements down the column of each sector j are the different components of the sector s upstream supply chain, the requirements from other sectors to meet its final demand, and the sum of the elements down the column is the footprint of sector j. 92

93 The SRIO variant of the above is detailed in Chapter 2 of this thesis. However, it simply involves assuming that r = s = 1 in the above system. For purposes of comparison between IRIO results for the UK and moving to the Scottish IRIO analysis, we focus on the own-country (i.e. r = s) results for equation (5) Emissions hot-spots In the first chapter we established a methodology to identify hot-spots in downstream and upstream supply chains. There are three types of hot-spots : (a) A sector that in producing output generates directly significantly more emissions compared to other sectors in an economy either to support total final consumption demand or components thereof; i.e. has a larger sum of its row in CO 2 emissions matrix (5); (b) A sector where the output produced to meet its final demand (again, either in total or components thereof), directly and/or indirectly, has a larger footprint, i.e. larger sum down its column in (5), compared to other sectors in an economy; (c) A point in a sector s downstream or upstream supply chain, an element within (5), that embodies emissions above a set threshold level in serving all or particular type(s) of final consumption demand. The same methodology is also used in this chapter to identify the hot-spots in Scottish supply chains. The sum of the elements on each sector s row in (5) is calculated and the sectors with significantly higher sum are classified as Type (a) hot-spots. In the same way the sum of the elements down the column of each sector in (5) is calculated and the sectors with the largest sum are considered as Type 93

94 (b) hot-spots. Finally, for the purposes of this chapter, and this thesis as a whole, the average of the row maximums in (5) is set to be a threshold level, above which every element of (5) is considered as a Type (c) hot-spot Data This chapter employs two distinct databases. For the IRIO analysis, focusing on the UK as a case study therein, the pilot OECD Inter-country Input-Output Database (OECD, 2015) 14 is used focusing on the most recent data of The database consists of 57 countries, both OECD and non-oecd members, plus a composite Rest Of the World region (see Appendix 3.A for a full list of countries); Industrial sectors have been grouped into 37 sectors following ISIC v3.1 (see Appendix 3.B for complete list of sector groupings); 5 sources of final demand. For the SRIO the 2009 Scottish Industry by Industry IO tables have been used (The Scottish Government, 2014). The reason is to match the year of the IRIO so that the findings could be compared. The Scottish IO tables include 98 industrial sectors; 8 types of Scottish final demand; 2 types of exported final demand, including Rest of UK and Rest of World. 14 In this chapter an earlier version of the database is used. 94

95 The data in the Scottish IO tables were reported in GBP therefore to match the data in the OECD Inter- Country Input Output Database it was necessary to convert them to US Dollars using the average exchange rate of 2009; 1GBP = USD. Also in the interest of enabling comparisons the emissions account used in the IRIO and also in the third chapter was also used here. We start out by applying UK intensities, e 0, for each sector in the Scottish system (and later introduce Scottish-specific data). This allows us to consider the impact of focusing on the region-specific economic interactions that are important in the Scottish case, abstracting from the region-specific polluting characteristics. Mapping from the UK data within the OECD emissions dataset ensures that the same types of emissions are allocated to each group of sectors (i.e. emissions from fuel combustion, autoproducer emissions, fugitive gases, industrial processes). However, since the Scottish IO tables use UK Standard Industrial Classification of economic activities 2007 (SIC 2007), while the OECD database used the International Standard Industrial Classification of all economic activities rev. 3.1 (ISIC rev. 3.1), it was necessary to match the OECD sectors with the Scottish IO sectors. Appendix 3.B shows how OECD sectors are matched to Scottish IO sectors. Additionally, because of different levels of aggregation, it is assumed that multiple Scottish IO sectors that belong in the same group under OECD classification, share the same CO 2 emissions intensity. There is a single case where three UK sectors under OECD classification are included in one Scottish IO sector and in this case the CO 2 emissions intensity used is the average of the intensities of the OECD sectors. Appendix 3.C shows the CO 2 emissions intensities of the Scottish sectors. The initial analysis then in Section 4.4 is conducted using the same data, and therefore the same emissions account, that was used in the previous chapter. In Section 4.5 we consider the impact of introducing Scottish specific emissions-intensity data for key polluting sectors, an approach that has 95

96 been followed by McGregor et al (2008). There we draw on data produced by the Scottish Environmental Protection Agency (SEPA) 15 as published on their website as well as data from the Longannet power station EMAS Since Electricity is found to be a rather important Scottish sector in terms of direct CO 2 emissions, it is necessary to have a clear picture on the active thermal power plants in Scotland. A list of the currently operational thermal power stations in the UK can be found at the Gov.uk website Case study demonstrating the benefits of using SRIO based on national/regional economic IO data to add information to an IRIO hot-spot analysis As explained in the introduction, the Scottish regional economy within the UK national economy is used in this chapter as a case study. The process that will be followed in this section is to first present and discuss results of an IRIO hot-spot analysis focused on UK sectors. These results are then compared to the results derived from a SRIO analysis based on the published Scottish IO tables. The goal of this section is to demonstrate in a clear way the additional information that can be obtained by using national/regional IO tables when focusing on a single country/region. Our first point of focus, which is important going forward, is direct emissions. By analysing IRIO data for r =UK as the producing region, we identify the total direct emissions of UK sectors. Table 1 shows the top 10 UK sectors with the largest volume of direct emissions, i.e. the largest sum of elements along their row in (5)

97 The Total Direct Emissions column (first column of results) shows the sum of the elements along the row of each sector in (5) and the second column presents them as a percentage share of the total UK direct emissions. The third column refers to the element e X 0 of each sector in (3) for r =UK. The fourth column is the total final demand of each sector, y V 7 in (2) and (5), while the last column is the sum of the elements along the row of each sector in (2). From the data in Table 1 it can be seen that the sectors that have the most significant share of UK s total direct emissions are Electricity, Gas and Water Supply (hereafter EGWS), Transport and Storage and Wholesale and Retail; Repairs. These three sectors can be classified as Type (a) hot-spots. However, they all share a common characteristic. As can be seen in Appendix 3.B, all three sectors are aggregated groups of industries under ISIC rev3.1. The gives rise to a problem in that it is not possible to make a distinction on which part of the aggregated sectors is generating the majority of the direct emissions. For example, Transport and Storage is the second largest Type (a) hot-spot amongst UK sectors. However, transportation of passengers and cargo can utilize numerous means, all of which are likely to have different environmental impacts in practice. If data are only available on the aggregated sector it is not possible to identify which of the means of transportation is generating the largest share of the aggregated sector s direct emissions. From a policy perspective this lack of information limits the 97

98 information set available to policy-makers and, thus, their ability to target policies. It is not possible for policy makers to identify whether the main contributor to the aggregated sector s total emissions is road transports, railroads, marine traffic, aviation or even the transportation support activities (e.g. warehouses, travel agencies etc.). Under this information constraint the only possible approach is to implement policies that involve all of these activities, even if most of emissions are generated by only one of them. Therefore, any implemented policies may be somewhat broad-brush rather than well targeted/focused. A different aspect of the over-aggregation problem is demonstrated by the third largest Type (a) hotspot. Data on Table 1 show that the emissions intensity of Wholesale and Retail; Repairs is amongst the lowest in all the sectors of Table 1. Therefore, the volume of direct emissions generated is due to the volume of output involved, which is the second largest amongst the Table 1 sectors. Examining columns Total Final Demand and Total Output for Wholesale and Retail; Repairs it can be seen that more than half of the sector s output is produced to meet the demand from final consumers (i.e. not the wholesale part of the sector). Having a large aggregated sector makes it impossible to identify exactly which part of the sector s activity is serving the largest share of the aggregated sector s final demand so that different demand drivers of associated emissions cannot be clearly identified. In both cases having additional information and more details on sectoral level interactions can aid in making more informed decisions. By being in a position to identify exactly which part of an aggregated sector holds the largest share of the sector s emissions, enables the introduction of policies that focus directly on the activities of this specific part/sub-sector. The potential benefit from such an approach is to use the available resources in a more efficient way, achieving the desired results faster and with reduced costs compared to trying to change an aggregated sector. 98

99 To that end, disaggregated regional and national statistics 17 can prove to be a valuable source of data. Table 2 shows the top 10 Scottish sectors in terms of direct emissions, i.e. the sum of the elements along the row in (5) for r = s =Scotland. As a reminder the disaggregated Scottish sectors have the same emissions intensities as the aggregated UK sectors. For the matching between UK and Scottish sectors please refer to Appendix 3.B. Similarly to Table 1, the first column of Table 2 is the sum of the elements along each sector s row in 5 while the second column shows the emissions of the first column as percentage share of the total Scottish direct emissions. CO2 Intensity column shows the element e X 0 of each sector in (3) for r =Scotland. Finally the Total Final Demand column shows the y V 7 for each sector in (2) and (5) and for s =Scotland, where Total Output is the sum of the elements along the row of each sector in 2. Note that the activity level results reported for Scotland in Table 2 are significantly smaller the UK equivalents in Table 1. This is broadly in line with the fact that Scottish sectors produced around 8.4% of the total value of output produced in the UK. At the same time the calculations performed using the Scottish SRIO indicate that the total direct emissions of Scottish sectors are roughly 11.5% of the total 17 As discussed during the literature review it has been show by Lahr and Stevens (2002) that highly aggregated regional and national IO tables have significant errors as a result of the aggregation level. Our goal though is to demonstrate the benefits of using disaggregated tables, therefore, from this point onwards whenever regional and national IO tables are mentioned it is implied that they are disaggregated unless stated otherwise. 99

100 direct emissions of the UK sectors. However, the key benefit of IO analysis is that it allows us to examine the composition of activity and this is what Table 2 allows us to begin doing. In Table 2, 4 of the top 5 sectors ( Electricity, Waste, Remediation & Management, Water and Sewerage and Gas etc. ) are all disaggregated sub-sectors of the IRIO Electricity, Gas and Water Supply (EGWS). Thus, it can be seen that using disaggregated regional IO tables such as the Scottish IO tables allows for the disaggregation of highly aggregated sectors such as EGWS. This is a fitting illustration of the discussion above on how disaggregation allows researchers and policy makers alike to clearly view that the main share of the emissions of Scottish EGWS is generated by Electricity. In fact, Electricity as a sector has the largest share amongst all Scottish sectors, 35.49% of the total Scottish direct emissions. Under a full EGWS aggregation the output of the analysis would indicate that Scottish EGWS generated Mt of CO 2 and would draw the attention equally to all of its components, whereas the data in Table 4 clearly show that Electricity generation requires increased policy attention, in the context of CO 2 emissions, compared to Water and Sewerage. Another example, related to the observations from Table 1, comes from the examination of the sectors ranked 8th and 9th in Table 2. As discussed above, Wholesale and Retail; Repairs is a Type (a) hotspot for the UK and the main driver of the sector s emissions is the sector s total output, more than half of which involves the production of final goods and services. Scottish sectors Retail-excl vehicles and Wholesale-excl vehicles are both parts of the aggregated UK sector Wholesale and Retail; Repairs. By using the Scottish IO tables, it is possible to identify that Retail-excl vehicles is generating more direct emissions but also that the majority of the emissions are generated to meet the sector s final demand, as the value of the final demand is almost the same as the sector s total output. From a policy point of view, this finding indicates that in order to reduce the emissions generated by Retail-excl vehicles it might be preferable to examine the consumer behaviour and trends. Additional reflection will be required to identify ways in which changes to consumption patterns can be achieved with the minimum 100

101 losses possible in terms of employment or value-added. On the other hand, the data in Table 2 show that there needs to be limited attention on the emissions generated by the Retail-excl vehicles sector, as its emissions intensity is significantly lower that any of the sectors in Table 2. In general, comparing the data in Tables 1 and 2 it can be seen that the Scottish sectors included in Table 2 make up the aggregated UK sectors of Table 1, or at least they are part of the sectors in Table 1. It can be seen then that by using regional IO tables we obtain more information on which parts of each aggregated sector have the most significant environmental impact. Regional IO tables also have a rather important advantage when the focus is on the identification of Type (c) hot-spots in supply chains. Table 3 shows the Scottish type (c) hot-spots where the producing sector is also included in Table 2. In Scottish SRIO the threshold level for an element of (5) to be considered a Type (c) hot-spot is 0.74 Mt of CO As shown in the previous chapters the threshold for Type (c) hot-spots is set to be the average of the row maximums in the CO2 emissions matrix. Therefore for the Scottish SRIO it is the average of the row maximums of (5) when calculated using the published Scottish IO tables. 101

102 The Embodied Emissions column of Table 3 shows the element of (5) where i = Producing Sector and j = Supporting the Final Demand of Sector. In the other two columns, the emissions of the first column are presented as a percentage share of the total direct emissions of the producing sector and the total direct emissions of Scottish industries, respectively. As can be seen in Table 3 the majority of the hot-spots account for a large share of the total direct emissions of the producing sector in question. Moreover, in some cases, e.g. the production of Electricity and Coke, Petroleum & Petrochemicals to meet own-sector final demand, the hot-spot also accounts for a significant share of the total Scottish direct emissions. It can also be seen that the majority of the Type (c) hot-spots in Table 3 involve the production of sectors to meet their own final demand. There are, however, two key exceptions, the production of Electricity to support the final demand of Mining Support and the production of Gas etc. to support the final demand of Electricity. The latter shows one of the advantages of using regional IO tables. Looking for UK Type (c) hot-spots on the IRIO, the largest one is the production of UK EGWS to meet its own final demand. With EGWS being an aggregated sector it is not possible to distinguish between (a) what share of the emissions embodied in this hot-spot involve output of one sub-sector of EGWS that is used as an intermediate input by another, and (b) which share is generated in directly serving final demand in any one subsector. Using the Scottish IO tables, which have EGWS disaggregated into 4 sectors, it can be seen that a significant part of the emissions generated by Gas etc. (16.66% in Table 3) is supporting the final demand of Electricity. However, the use of regional IO tables has failed to provide a solution to one of the problems highlighted earlier in this section. We identified that Transport and Storage, the second largest UK Type (a) hot-spot, involves different means of transportation and support services, each with different environmental impact. By using the UK emissions intensities for the Scottish SRIO, the problem is carried through to the disaggregated sectors like Support Services for Transport and Other Land Transport 102

103 that are included in Table 2. These sectors are part of the UK Transport and Storage sector in the IRIO and as a result of not using region-specific emissions data, they are assigned the UK emissions intensity. However, the Support Services for Transport sector includes activities such as the handling of cargo and the operation of travel agencies, which involve significantly less use of fossil fuel and therefore CO 2 emissions intensity compared to Other Land Transport activities. In short, it would seem then that the use of the same emissions intensities for the IRIO and the SRIO that uses regional IO tables leads to errors in terms of the calculated emissions. 4.5 Added value from introducing region-specific data for Scotland Emissions reported by Scottish industries The Scottish direct emissions presented in Table 2 in the previous section were the result of calculations performed using the Scottish-specific economic component but UK emissions intensities from the OECD framework in a SRIO analysis. From those calculations it was found that Electricity and Coke, Petroleum & Petrochemicals are the sectors that generate the majority of emissions in Scotland and actually Electricity has a rather large share of the total Scottish direct emissions. It is interesting then to see how the calculated figures compare to the ones reported to SEPA for our accounting year (2009). Starting with the Electricity sector, according to Gov.uk (2015) there were 10 thermal power plants operational in Scotland at the time of reporting. Of these the only ones currently included in data reported by SEPA are the coal-fired Longannet, the CCGT plant in Peterhead and the diesel-fired plant in Lerwick. The remaining plants are minor producing units located in Scottish Isles and mainly cater for the needs of local residents. However, for our accounting year of 2009 though there were two more power plants in operation and reported to SEPA. The one was the coal-fired plant Cockenzie and the other the CCGT plant in Fife. 103

104 In the situation for 2009 analysed in the SRIO work above, the level of CO 2 emissions reported by the 5 plants included in the SEPA data was approximately 13.4 Mt of CO 2. Clearly this figure is significantly smaller than the total we have calculated using UK intensities (from our OECD dataset) applied to the Scottish SRIO analysis, which was found to be 37.3 Mt of CO 2 in Table 2. Given the size of the remaining thermal power plants, the emissions left unreported should not be enough to fill the gap between what is calculated for Scotland using UK intensities and the reported emissions for Scotland in A reasonable explanation on the observed difference is that the emissions intensity of the Scottish Electricity sector is significantly smaller that the intensity of the aggregated UK sector EGWS. In fact, the mixture of electricity producing installations in Scotland suggests that this is the case and that the Scottish Electricity sector is likely to be less CO 2 -intensive than its UK counterpart (which is hidden within the EGWS sector above). Scotland has a large number of wind farms, while there are also hydro plants and biomass units. This essentially means that a relatively large share of Scottish electricity is being generated without the emission of CO 2 and therefore the sector s emissions intensity is reduced (Turner et al, 2011b). The Scottish Electricity sector still contributes approximately 10% of the total value of electricity produced in the UK, but with reduced embodied emissions compared to the rest of the UK. Despite being slightly irrelevant, given the focus of this chapter, we should note in considering recent industrial emissions data a key limitation of IO analyses (either SRIO or IRIO). This is that due to their complexity economic IO tables tend to be reported with a time lag that may be important when we are considering an issue such as emissions generation in a rapidly changing energy supply environment. Scottish IO tables are now available for 2012 and the most recent year with reported emissions in SEPA is 2013 so that the SRIO analysis conducted here could be updated. However, for the purposes of illustrating the move from and comparability between IRIO and SRIO analyses of emissions generation, we continue to focus on our accounting year of However, in doing so we should note some key 104

105 changes that have occurred between now and then. For example, by 2013 the CCGT plant in Fife was no longer operational and even though the emissions by Longannet and Lerwick were significantly increased, drops in emissions by Peterhead and especially Cockenzie, which actually was decommissioned in 2013, meant that the emissions total from Scottish thermal plants in 2013 was down to 11.3 Mt of CO 2. Emissions for 2014 have not been reported but according to Longannet s EMAS for 2014 the level of emissions was slightly reduced relative to Another Scottish sector that has distinct characteristics in terms of CO 2 emissions is Coke, Petroleum & Petrochemicals. This Scottish sector partially maps to the OECD-defined Coke, Refined Petroleum Products and Nuclear Fuel but also includes petrochemicals industries that in the IRIO table are part of the OECD-defined Chemicals and Chemical Products sector. However, since we have not enough data to be able to identify which part of the Chemicals and Chemical Products output involves the production of petrochemicals, the results reported above are generated under the assumption that Coke, Petroleum & Petrochemicals includes the same industries as the OECD-defined UK sector Coke, Refined Petroleum Products and Nuclear Fuel. There are a number of industries in Scotland belonging to this sector and in 2009 their CO 2 emissions as reported in the SEPA data were 5.6 Mt of CO 2. The data from SEPA suggest that the analysis (using UK intensities from the OECD data) overestimated the emissions of that sector. The direct emissions reported in Table 2 are around 9.5 Mt of CO However, in this case there is an important detail that must be taken into account. The UK CO 2 intensity derived from the OECD data also accounts for the operation of CHP autoproducing plants in industries. The emissions of such plants are not reported as part of the emissions generated by the industries of Scottish sector Coke, Petroleum & Petrochemicals. They are instead reported for thermal power plants. However, since they are not connected to the grid and are used for the needs of specific industries, these emissions are not captured by the calculations performed for the Electricity sector 19 The CO2 emissions of petrochemicals industries are just Mt of CO2 105

106 either. To avoid having such sources of emissions left unaccounted for, the practice used in constructing the OECD data set was to allocate them to the industries where they are actually used to facilitate the production. The amount of emissions associated with autoproducing plants is not negligible. For example, in 2009 the CHP plant that operates in Grangemouth generated approximately 0.7 Mt of CO 2. Correcting for this issue will then close the gap between the SEPA figure and what is calculated (using UK intensities) for the Scottish Coke, Petroleum & Petrochemicals sector. When the emissions of the CHP plant are accounted for, the calculated emissions - using the Scottish SRIO and UK intensities- are still 3.2 Mt of CO 2 larger that the emissions reported to SEPA The importance of region specific satellite emissions data As has been shown for both Scottish sectors Electricity and Coke, Petroleum & Petrochemicals the emissions calculated by using the UK intensities from the OECD IRIO system in Scottish SRIO led to overestimations compared to emissions actually reported for Scotland in the same accounting year of The most obvious implication is that the direct emissions allocated to both sectors were larger than the actual emissions generated. However, the implications do not stop there. In an IO analysis the direct emissions of one sector impact the multipliers and emissions attributable to activity in other sectors. To consider the wider impacts that even one of these overestimations has across the SRIO for Scotland, we use the SEPA reported emissions for Electricity sector in recalculating the sector s CO 2 intensity. Given the problems discussed above in allocating emissions to the Scottish Coke, Petroleum & Petrochemicals sector (both in terms of using UK-intensities and the problems of accounting for CHP plants), we do not attempt to introduce region-specific data for this sector. Rather, we focus attention on the impact of making a single adjustment for region-specific data in the key polluting sector of Electricity. We recalculate the SRIO system with just the single change of introducing the SEPA Scottishspecific data for e 0 where i = Electricity. 106

107 In Table 2 we found that the majority of the emissions generated by Electricity sector are used to support the final demand of other sectors rather than meet the sector s own final demand. This means that, apart from the difference in Electricity direct emissions, there is a difference in how the emissions intensity of Electricity impacts the Type I emissions multiplier for each sector j, i.e. the sum down the column of each sector in (4), for every Scottish sector. As a result, when we recalculate there are differences in the footprints of Scottish sectors, with major differences being observed, as would be expected, in sectors that require large amounts of Electricity output. Table 4 shows the new results using the adjusted (for region-specific data) Scottish Electricity intensity. We report results for the top 10 Scottish sectors in terms of footprint (Scottish emissions attributable to final demand for sectoral output), presenting these alongside a summary in the same format for what we found using the UK intensity from the OECD IRIO system. In Table 4 the column Total Emissions (Footprint) shows the sum of the elements down each sector s column in (5) while the column Type I Emissions Multiplier is the sum of the elements down each sector s column in (4). As can be seen from the data in Table 4, adjusting the CO 2 emissions intensity of Electricity has changed the footprint of every sector and changed the ranking. The most significant change was on the footprint of Electricity itself, which was expected given that the data from Tables 3 and 4 show that the 107

108 most significant contributor to the pre-adjusted Electricity footprint was the sector s own production. Therefore, the adjustment of Electricity CO 2 emissions intensity, which is significantly smaller than the one in the IRIO, led to a major reduction to the sector s own footprint. The identity of the other 9 of the top 10 sectors in terms of footprint are exactly the same in both cases. However, it can be observed that there is a (downward) change in terms of the magnitude of the footprint and also in the ranking of the sectors. The most significant changes are the Wholesale excl vehicles being ranked 6th (it was originally 10th) after the adjustment and Mining Support being ranked 10th after the adjustment when it was 6th originally. Especially in the case of Mining Support the reduction in the sector s footprint is 0.86 Mt of CO 2, the second largest after the reduction in Electricity footprint. This result may have been expected given that the output of Electricity used to support the final demand of Mining Support is a Type (c) hot-spot, as can be seen in Table 3. In fact, adjusting the CO 2 intensity of Electricity has reduced the emissions embodied in the Electricity output produced to support Mining Support final demand enough to be below the threshold level for Type (c) hot-spots, even though the threshold level was also reduced to 0.57 Mt of CO Conclusions and extensions The goal of this study was to demonstrate that the use of disaggregated national/regional IO tables can provide additional details on the results found by using IRIO tables. UK and Scotland were used as a case study. It was shown that the aggregated sector EGWS is a major UK Type (a) hot-spot and that the sectors that are included in EGWS are Type (a) hot-spots within Scotland (the trade-off in moving to SRIO is that which provides argument for IRIO in the literature not being able to account for impacts of international trade, or even interregional UK trade here). However, disaggregation at the Scottish level has revealed that it is in fact the Electricity component that has the largest share of EGWS emissions in Scotland and therefore is the largest Scottish Type (a) hot-spot. Furthermore, it was revealed that 108

109 there are significant emissions generated by Gas etc. to support the final demand of Electricity, which would otherwise be included in the EGWS production to meet its own final demand. However, there have been observed significant issues that need to be addressed in future projects. The most important one is that it is required to use region specific emissions data where possible. Using common emissions intensities at both regional and country level does ensure that each regional sector has been allocated the same emissions with the sectors at the country level (for example petroleum refineries attributed emissions also include the emissions of autoproducing plants). However, lack of region-specific information could lead to major errors in both PAP and CAP calculations. Particularly in sectors such as Electricity, where there is a wide variety of generation technologies, the use of nonregion specific emissions data can lead to significant overestimation or underestimation of the total emissions. The fact that potential overestimations might occur was uncovered by retrieving the reported emissions of specific sectors, while it was demonstrated that using the reported emissions to adjust the emissions intensity of Electricity led to a reduction of footprint for all the Scottish sectors. The findings of this chapter are in line with the findings of Turner (2006) regarding the benefits from using region specific data to study the environmental impact of economic activities in Jersey. Arguably then, devolved governments with powers over environmental targets and regulations like Scotland, would have a strong incentive to promote the development of regional emissions data to be used with their existing IO tables. The ability to implement targeted/focused policies is even more important as the available resources for the reduction of emissions will potentially be less compared to the national/federal government. However, as was discussed at the conclusion of the previous chapter, even though Scottish Government has a dedicated department that develops and publishes high quality IO data, there has been limited emissions related application. Furthermore, in some cases UK data were used instead of Scotland specific. The findings of this chapter could then act as an argument in favour of publishing Scottish emissions data for each sector in the Scottish IO tables. Especially since Scotland has 109

110 set ambitious emissions reduction goals (Climate Change (Scotland) Act 2009 found at: Legislation.gov.uk, 2016), being in a position to identify the sectors where policy interventions will have the most significant impact could prove to be rather helpful in meeting both the interim and the 2050 target. What was mentioned above though does not mean that the only benefited parties from detailed emissions data are regional governments (e.g. Scottish Government). However, national governments (e.g. UK Government) are more likely to have the resources and the necessary funds to develop both the disaggregated IO tables and the satellite emissions accounts. This issue has been highlighted by Turner (2006). There is a significant resource cost in developing and publishing region specific emissions data, which is higher the larger the economy. The cost will be larger for Scotland, being a much larger economy than Jersey. In that sense, an important factor that determines whether there should be investment in publishing region specific data is the gap between the expected benefits of those data against the associated cost of development. It is not the scope of this thesis to answer this question, which could be the subject of a different research project. However, given the significant differences observed as a result of introducing Scotland specific data in a single but important sector, it is our belief that region specific data need to be developed, at least to an extent that can be used to gauge their benefits against their costs (Turner, 2006). Reflecting on the findings of this chapter in association with the conclusions of the previous chapters, as well as remarks from the literature, inevitably the issue of the quality of the data is brought forward. As mentioned before, the Scottish IO tables are produced by a dedicated department of the Scottish Government and can be trusted on the quality and the accuracy of the tables they publish. The same cannot be said for the CO 2 emissions. It was observed that the reported emissions were mostly calculated by industries using SEPA approved methods. Although this is a common practice in calculating and reporting the emissions generated for example due to combustion of coal, obtaining actual physical measurements data would be beneficial in our understanding of the impact of specific sectors, in terms 110

111 of CO 2 emissions. Keeping these parameters in mind a future step for this project would require the acquisition of Scottish emissions data that, where possible, is suitable for following the standards of emissions satellite accounting used in the second chapter. Using the region-specific data would then make it possible to identify again the highest emitters with a higher degree of accuracy and to use findings of IO hot-spot analysis as a guideline to help focus policy and industry attention on the need for better and more updated physical measurements on which to focus current policy attention. Such an approach will help policy-makers better understand the nature of present production conditions at the location of industries belonging to those sectors with the IO analysis (both SRIO and IRIO) adding insight in terms of demand drivers, key supply chain relationships etc. 111

112 Chapter 5: Thesis Conclusion As it has been discussed throughout this thesis, Input Output (IO) has been used extensively to calculate the CO 2 emissions generated by economic activities. Researchers have used both Single Region IO (SRIO) and Inter-Regional IO (IRIO) to measure the emissions generated under a Production Accounting Principle (PAP) and a Consumption Accounting Principle (CAP). The differences in allocated emissions under each of the two principles has been the subject of different studies, given that international climate change mitigation agreements, like the UNFCCC agreements, utilise a territorial PAP to assign responsibility for the emissions generated to each participating party. There have been arguments against this approach and in favour of CAP claiming that PAP does not capture the emissions embodied in international trade correctly (e.g. Wyckoff and Roop, 1994; Munksgaard and Pedersen, 2001) or pointing out that PAP could lead developed countries in relocating their most polluting activities in developing countries with more lenient environmental regulations, i.e. carbon leakage (Arrow et al, 1995). We believe that both PAP and CAP can contribute significantly in our understanding of the drivers of emissions. However, the way these headline figures are calculated there is limited transparency on the structure of the emissions; PAP and CAP do not provide any information on the share of emissions embodied in each of components of any sector s downstream and upstream supply chains, the role of each sector s emissions intensity or the impact of final demand. Recognising these limitations of working with headline figures, our project aimed to develop a methodological tool that makes use of the information provided by PAP and CAP while providing transparency in their calculation, which most of the time has the characteristics of a black box. Motivated by this limited transparency on the structure of emissions and the lack of information on the impact of emissions intensity and final demand on the total volume of generated emissions, we developed what we call a hot-spot analysis method. Hot-spots were identified by initially using the headline figures of total direct emissions and CO 2 footprint, which we called Type (a) and Type (b) hot- 112

113 spots ; they reflected the emissions embodied in each sector s downstream and upstream supply chains and they are part of the PAP and CAP emissions of each sector. Through the analysis of direct emissions, it became possible to understand whether the volume of emissions is driven by the sector s emissions intensity or the total volume of output, while the comparison between the total output and the final demand helped distinguish between the sectors that mainly produced to support other sectors and those that produced to meet their own final demand. In a similar way, analysing the CO 2 footprint of different sectors helped identify sectors that have significant emissions embodied in their upstream supply chain due to the volume of their final demand and others that their emissions were driven mainly by their Type I emissions multiplier, a feature of our method that builds on the estimation of backward linkages using Type I output multiplier that was established by Rasmussen (1957) and Hirschman (1958). We then used the Type (a) and Type (b) hot-spots we found as a guideline to identify downstream and upstream supply chains that would be interesting to disaggregate and look for Type (c) hot-spots in components of these supply chains. This is one of the key benefits that we believe that can be gained by using hot-spot analysis. We were able to highlight the components of downstream and upstream supply chains of different sectors, both within the same and across different regions, that hold the largest shares of the total embodied emissions in the supply chains. Putting this feature into policy perspective it means that policy makers who have identified important sectors in terms of being large direct emitters and/or having a large CO 2 footprint can focus their attention, policies and resources on Type (c) hot-spots on the supply chains of those important sectors, rather than implement policies on those sectors as a whole. Furthermore, we were able to analyse the Type (c) hot-spots and reveal which, amongst the emissions multiplier or the receiving sector s final demand in the case of hot-spots on the downstream supply chains; and the emissions intensity and the requirements of the consuming sector from the producing sector in the case of hot-spots on upstream supply chains, was the main driver of the embodied 113

114 emissions in these Type (c) hot-spots. This is similar to the analysis of the Type (a) and Type (b) hotspots and the other key benefit of hot-spot analysis. Policy makers not only have the ability to identify where the major shares of emissions are located on the sectors supply chains, but they are also in a position to understand how different parameters influence the embodied emissions and which one of them is the main driver. All these features of our method make for a more complete set of information regarding the emissions generated by different industrial sectors and making use of this information can eventually lead to more informed and better targeted/focused policies, as discussed throughout this thesis. In the different chapters of this thesis the proposed methodology has been applied to different IO frameworks ranging from a SRIO which was derived by an IRIO table, hence it had the same level of sectoral aggregation, to a full IRIO to a disaggregated sub-national IO dataset published by the Scottish Government. Applying the methodology to a range of IO frameworks offered us the opportunity to develop the methodology but at the same time identify its limitations which can be addressed in future research projects associated with this thesis. In the second chapter the methodology was applied to a SRIO for China in 2005, which used the data from the Chinese component of the OECD Inter-Country Input Output database. Conducting hot-spot analysis led to the identification of a number of Chinese sectors as hot-spots based on their direct CO 2 emissions, i.e. Type (a) hot-spots. The largest Type (a) hot-spot was found to be the aggregated Electricity, Gas and Water Supply (EGWS) sector, which generated more than half of the total emissions directly generated by Chinese sectors. However, further analysis of the sector s downstream supply chain has shown that the majority of the emissions were generated to support the final demand of other Chinese sectors, mainly Construction, while the emissions generated to meet its own final demand accounted for just 9.7% of EGWS total direct emissions. Analysis of all Chinese industries (as defined in the OECD database) has shown that Construction is the largest hot-spot in terms of 114

115 domestic footprint of production to meet final demand (i.e. Type (b) hot-spot ), mainly in the form of domestic capital formation. This is an interesting finding, given that capital formation in 2005 will have played a key role in facilitating economic development, including increased production to meet export demands over the subsequent decade. Additionally, hot-spot analysis of Construction upstream supply chain has highlighted a number of components of the Construction upstream supply chain that can be classified as Type (c) hot-spots and which are also the most polluting outputs of the respective sectors, e.g. the output of Mining and Quarrying and Coke, Refined Petroleum and Nuclear Fuel to support Construction final demand. As was discussed in the chapter, our findings indicate that if policy makers in China wish to reduce the embodied emissions in the EGWS downstream supply and the Construction upstream supply chain, it would be worthwhile to address the Construction requirements for EGWS output. Having the information from hot-spot analysis available means that policy makers are aware of the different options; to either re-structure EGWS to achieve a lower emissions intensity or reduce the Construction requirements for EGWS output by introducing more energy efficient building technologies. Unfortunately, as pointed out throughout this thesis, hot-spot analysis is not a stand-alone tool that can be used to conclusively decide which of these two options is preferable. Additional analysis on the employment and value-added impacts would be necessary to reach to valid conclusions, while a more sophisticated modelling tool like Computable General Equilibrium (CGE) would be crucial in simulating the impact of any policies over a longer period of time. However, hot-spot analysis still reveals that the volume of embodied emissions in the supply chains of EGWS and Construction heavily depend on the interaction of the two sectors and policy makers can focus on this interaction rather than implement policies on each of the sectors separately. Apart from the fact that hot-spot analysis is a specialised rather than a stand-alone analytical tool, in the second chapter a few more limitations of the methodology were identified. A basic sensitivity 115

116 analysis conducted in section of the second chapter showed how important it is to set the Type (c) hot-spot threshold correctly. A small variation in the threshold level could mean that several components of different supply chains will be considered or disregarded as Type (c) hot-spots. Looking back at Tables 3 and 6 in Chapter 2 and 3 and 5 in Chapter 3, it is noticeable that in every one of these tables there are sectors included, which have embodied emissions near the threshold level. In much the same way there are sectors in the CO 2 emissions matrix that have embodied emissions just under the threshold level and are therefore excluded from the tables. Even though this is a characteristic of every type of system that uses a cut-off point to distinguish between important and less important entries, this is a feature that needs to be taken into account when conducting hot-spot analysis. The fact that a sector generates emissions just below the threshold level does not imply that it does not merit any policy attention and, more importantly, that it does not offer the possibility to achieve reduction in the country s or region s total emissions with a relatively small cost. To solve this issue, it was suggested that the threshold level can be set based on the average embodied emissions required per sectoral transaction for each country to meet its targets set either voluntarily or by international agreements. Alternatively, a confidence level type of approach could be implemented by using the difference between the threshold level (which we set at the average of the row maximums) and the smallest Type (c) hot-spots. Any components of supply chains with embodied emissions below the threshold level but within that confidence level could be considered as Type (c) hot-spots. This is purely a suggestion of what could be done in future projects that was not applied to this project as any calculations were done for illustration purposes, to show how the proposed methodology can be used, highlight its strong points and reveal its limitations. The other limitation that was highlighted in Chapter 2 involves the application of the hot-spot analysis on single, discrete years. Economies are dynamic systems and as such they change over the years. By examining a static picture of an economy does not provide any insight on whether the identified hot- 116

117 spots are recurring or were one-off phenomena. Being able to make this distinction is very important to avoid the allocation of resources to sectors that for any reason generated or contributed to the generation of significant emissions at a single point in time and by the time of the analysis the problem has been resolved either by existing policies or due to changes in the structure of the economy. The lesson learned from considering this limitation is that historical hot-spot analysis would be necessary, when facilitated by the availability of data, to reinforce the robustness of the findings in the later years. In the third chapter the hot-spot methodology was extended and applied in an Inter-Regional Input Output (IRIO) framework, with a novel focus on decomposition of IRIO emissions accounting results, in order to allow the identification of hot-spots in global supply chains. In the applied work in the chapter, the focus was placed on considering the composition of the CO 2 footprint of UK total final consumption demand. Using the hot-spot methodology it was possible to identify a number of sectors outside the UK that directly generate significant volumes of emissions either to directly meet UK final demand or to support the UK final demand requirements of other sectors further up the global supply chain. The largest external Type (a) hot-spot serving UK final consumption demand was found to be the Chinese EGWS sector. Given that UK final consumers do not directly consume Chinese utilities services, as would be expected, examination of international side of the sector s downstream supply chain revealed that UK-supported hot-spots are located at points on the supply chain where the output of China s EGWS supports the UK final demand of other industries, primarily in the UK itself. This is an interesting finding on its own, one that flags up an important issue associated with all types of IO analysis. Given the nature of the outputs of EGWS, one might think that there is limited export potential from the Chinese EGWS to UK sectors. However, the findings show that at least some UK sectors do import significant volumes of inputs from the Chinese EGWS. This result act as a reminder that the quality and the accuracy of the results of every type of IO analysis are closely correlated to the quality and the accuracy of the IO tables used. UK s Department of Energy 117

118 and Climate Change (DECC) has raised the issue that there is lack of credible international data, especially for use on consumption-based approaches (House of Commons Energy and Climate Change Committee, 2012a). For this chapter, and this project as a whole, the OECD Inter-Country Input Output database has been used, which, alongside the transparency achieved by the proposed methodology, addressed the concerns of DECC over the availability of data while providing a step forwards regarding the quality of data. Even though we have the utmost confidence that the people who created this dataset have the highest standards, still the fact that only a pilot version of database was used, casts shadows on the credibility of results. Fortunately, the sole purpose of these results was to demonstrate the ways in which the developed methodology can be used and not provide conclusions to be used in a policy making setting. However, further applications of the hot-spot analysis methodology will need to be conducted in the final and most credible version of the data, which will not include any discrepancies. Another unanticipated finding was that the second largest Type (b) hot-spot was the UK Health and Social Work sector. It was found to have a significant number of hot-spots on the international side of its upstream supply chain. This is a finding that could have important policy implications. Particularly in public sector activity, actions to reduce the carbon footprint of health sector activities often focus very much on direct energy use, improved efficiency in buildings etc. That is, things that policy-makers have ready access to information on and a high degree of control over. However, the finding that the footprint of the UK health sector is impacted by emissions related to chemical production in other countries, while intuitive (given the international nature of industries such as pharmaceuticals), it raises challenges in terms of how these might be reduced. The chapter raised the issue of jurisdictional limits to what policy-makers in one country can do about emissions in another. However, it also highlighted the role that purchasing managers etc. both in the UK health sector and firms in their supply chains can play in applying commercial pressure on upstream suppliers to reduce their emissions. Furthermore, the findings of hot-spot analysis on international supply chains can be used, alongside a set carbon price, 118

119 as the basis to determine the mandatory contribution of each developed country to carbon reduction funds, which in turn will assist in the implementation of carbon reduction innovations in developing countries under the carbon finance concept. This is a good point to reflect on some of the issues that arise from applying the hot-spot analysis methodology on an IRIO. The identification of hot-spots on supply chains relies on setting a hot-spot threshold. It has been discussed so far the threshold level can be set to be the average intersectoral transaction required to meet the emissions targets of any country. However, different countries have different emissions targets, either set voluntarily or by international agreements. Setting then a common threshold level across all countries was useful for illustration purposes but does not reflect the differences in emissions goals. Keeping that parameter in mind, any future applications need to set different, country-specific, threshold levels and consequently the findings to reflect the differences between each country. The other issue is related to the very nature of IRIO tables. As was discussed in the previous chapters, the creators of the OECD database that we have used in this project had to harmonise data from various sources that do not follow the same approaches in collecting and reporting their data. As a result, it was necessary to aggregate sectors so that each country in the dataset has the same industrial classification, often leading to over-aggregated sectors like EGWS. However, as demonstrated by Miller and Blair (1981) and de Mesnard and Dietzenbacher (1995) amongst others, over-aggregation can lead to significant biases and errors. Therefore, hot-spot and any other kind of IO analysis would benefit from the use of disaggregated IO tables, at least at the national and sub-national level. The work in the fourth chapter focused on demonstrating exactly that; how it is possible to gain an additional insight on the findings from hot-spot analysis in an IRIO by moving to a more focused and sectorally detailed SRIO analysis. Building on the applied work in the third chapter, the UK was used as a case study for identifying hot-spots within the UK economy using IRIO methods and the OECD Inter- 119

120 Country Input Output (ICIO) database. Once the UK hot-spots were identified we then shifted our attention to examining the existence and nature of these at the Scottish level. This involved conducting the same type of analysis as was developed in Chapter 2 (there for the Chinese case study using an extract for that country from the OECD ICIO) but in the context of Scotland and using the regional IO tables published by the Scottish Government. This allowed us to work at a higher level of sectoral disaggregation. A key finding of the Scottish SRIO analysis was the revelation that Scotland does indeed share the same and the Scottish sectors are part of what we labelled as Type (a) or direct emissions generation hotspots for the UK using the IRIO. However, and with attention to the highly aggregated Electricity, Gas and Water Supply sector in the OECD ICIO, the Scottish findings reflected the concerns that have been expressed in the literature regarding aggregation issues. Using the more disaggregated Scottish IO system we found that what is identified a single direct CO 2 hot-spot in the UK IRIO analysis actually involves key sectoral interactions between the component Electricity and Gas etc. sectors in particular (use of gas in electricity generation). This is likely to be true at UK national level as well (indeed, in many countries) and emphasises the need to drill down on IRIO results to provide a fuller information set for policy analysts and decision makers. Another key finding of the analysis on the fourth chapter emerged from the examination of the emissions calculated using (a) region-specific economic data in the form of the Scottish IO tables but with UK emissions intensities reported in the OECD environmental satellite account (which the author was involved in the construction of) and (b) results when some region-specific environmental data were also introduced. We considered bottom-up industry level data on Scottish electricity generation reported via the Scottish Environment Protection Agency (SEPA). It was found that the system based on UK average industrial CO 2 intensities leads to false over-estimates not only of emissions directly generated in the Scottish Electricity sector but also in what we labelled as Type (b) and (c) hot-spots in 120

121 the overall and key upstream points in the CO 2 footprint estimates for sectors that directly and indirectly use electricity in their supply chains. Generally, we found that adjusting the emissions intensity for just one of the Scottish sectors (albeit an important one) it has an impact to the Type I emissions multiplier and therefore the footprint of all Scottish sectors. This finding and the one discussed previously highlight the importance of using region-specific economic and emissions data. Unfortunately, the resource cost of developing and publishing such data is a limiting factor, as highlighted by Turner (2006), especially since larger economies will have to face higher costs. However, the findings of our analysis suggest that there are benefits to be gained that merit investments in the development of such data, especially in countries and/or regions like Scotland that already have developed disaggregated IO tables. The example of Support Services for Transport and Other Land Transport in Table 2 in Chapter 4 is rather eye-opening. Without region-specific data the two sectors share a common emissions intensity as they both are part of the aggregated UK sector Transport and Storage from the IRIO. Following the findings in Table 2 without further consideration would suggest that Support Services for Transport merits more policy attention compared to Other Land Transport. However, it is a known fact that the dependence of Other Land Transport on fossil fuel is far greater than that of Support Services for Transport and therefore the two sectors should have had different emissions intensities. This can be achieved by introducing region-specific data, which as a result would lead to a higher degree of certainty that the sectors in Table 2 of Chapter 4 indeed directly generate the most CO 2 emissions in Scotland and hence require further investigation of their supply chains. In summary, this thesis has made important developments in terms of decomposing IO multiplier methods and applying these in the context of analysing hot-spots in domestic and global supply chains. It has also made an important applied contribution in terms of demonstrating the types of insights that can be gained through application of this methodology. However, even though every effort was put 121

122 through to produce the best results possible there is always potential to expand and improve the work presented in this thesis. For example, the IO datasets are constantly improving to reduce any discrepancies but also to include more countries and recent years. Therefore, a potential future project could conduct the analysis presented in this thesis using the most up-to-date datasets. Especially in the cases where more countries are included, analysis of the data could lead to the identification of additional hot-spots. Another potential future project could be to conduct physical measurements or use orbital measuring to obtain a picture of the present volume of emissions at the location of the Scottish industries identified as hot-spots. Such a project would involve combining the findings obtained by analysing the different supply chains with actual geographical data, so that a true spatial disaggregation of the supply chains is achieved. On a more practical application, several parties amongst which the House of Commons (2012a) have brought forward the concept of educating consumers, making it clear for them what is the economywide environmental impact of 100, for example, worth of private consumption/final demand. Our methodology can identify both the total impact and the specific locations in the domestic and global economy where this amount of final demand has the most significant impact. Finally, a key conclusion of the fourth chapter is that it is important to focus in on IRIO findings at national and sub-national regional level using more detailed economic IO data and to use, where possible, region specific emissions data. That being the case, if it becomes possible to obtain Scottish emissions data with the same characteristics as the ones used to create the OECD satellite emissions account in the third chapter, it would be rather interesting to conduct a full hot-spot analysis for the UK, Scotland and Wales (where region-specific IO tables are also available) and compare the findings. Where interregional trade data can be identified, it would also be interesting to apply IRIO methods to consider CO 2 hotspots linked to trade flows within the UK economy. 122

123 Acknowledgements This thesis is the result of a Ph.D. programme funded by ESRC-NERC (Grant Reference: ). I would like to thank ESRC/NERC for my scholarship as well as the Scottish Graduate School in Social Science. I would like to express my gratitude to my supervisors, Professors Karen Turner and Ian Simpson, for the offering me the opportunity to conduct this research and for their continuous advice and support both at personal and academic level. I would like to thank the examiners, both external and internal, for their constructive comments. I also want to thank the academic staff and the PhD candidates at the University of Stirling for their comments/suggestions during the departmental workshops. Special thanks are extended to the OECD Directorate of Science, Technology and Innovation and especially Mr. Norihiko Yamano and Mr. Colin Webb. Sharing the OECD database as well as ideas on how to tackle specific issues, offered invaluable assistance at the beginning of my research project and throughout. I would also like to thank my friends and family for their support. Last but most certainly not least I would like to thank my partner for her support and her understanding, especially at the times that my stress made me not very pleasant company. 123

124 References Acquaye, A., Wiedmann, T., Feng, K., Crawford, R., Barrett, J., Kuylenstierna, J., Duffy, A., Koh, S. and McQueen-Mason, S. (2011). Identification of carbon hot-spots and quantification of GHG intensities in the biodiesel supply chain using hybrid LCA and structural path analysis. Environmental science & technology, 45(6), pp Ara, K. (1959). The Aggregation Problem in Input-Output Analysis. Econometrica, 27(2), pp Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C., Jansson, B., Levin, S., Maler, K. and Perrings, C. (1995). Economic growth, carrying capacity and the environment. Science, 268(1), pp Bruckner, M., Giljum, S., Lutz, C. and Wiebe, K. (2012). Materials embodied in international trade Global material extraction and consumption between 1995 and Global Environmental Change, 22(3), pp Cadarso, M., Lopez, L., Gomez, N. and Tobarra, M. (2012). International trade and shared environmental responsibility by sector. An application to the Spanish economy. Ecological Economics, 83, pp Chenery, H. and Watanabe, T. (1958). International Comparisons of the Structure of Production. Econometrica, 26(4), p.487. Court, C., Munday, M., Roberts, A. and Turner, K. (2015). Can hazardous waste supply chain hotspots be identified using an input output framework?. European Journal of Operational Research, 241, pp

125 De Mesnard, L. and Dietzenbacher, E. (1995). On the Interpretation of Fixed Coefficients under Aggregation. Journal of Regional Science, 35(2), pp Dietzenbacher, E. (1992). The measurement of interindustry linkages: Key sectors in the Netherlands. Economic Modelling, 9(4), pp Ewing, B., Hawkins, T., Wiedmann, T., Galli, A., Ertug Ercin, A., Weinzettel, J. and Steen-Olsen, K. (2012). Integrating ecological and water footprint accounting in a multi-regional input output framework. Ecological Indicators, 23, pp.1-8. Gale, L. (2004). Trade liberalization and pollution: An input output study of carbon dioxide emissions in Mexico. Economic Systems Research, 16(4), pp Genty, A. (ed) (2012). Final database of environmental satellite accounts: technical report on their compilation, WIOD Deliverable 4.6, Documentation, downloadable at Gov.uk, (2015). Power stations in the United Kingdom. [online] Available at: s Hawdon, D. and Pearson, P. (1995). Input-output simulations of energy, environment, economy interactions in the UK. Energy Economics, 17(1), pp Hirschman, A. (1958). The Strategy of economic development. New Haven: Yale University Press. 125

126 House of Commons Energy and Climate Change Committee (2012a). Consumption-based emissions reporting, Twelfth Report of Sessions , Volume 1. HC 1646, published on 18 April 2012 by authority of the House of Commons. London. The Stationary Office Limited. Available to download at House of Commons Energy and Climate Change Committee (2012b). Consumption-based emissions reporting: Government response to the committee s Twelfth Report of Sessions HC 488, published on 25 July 2012 by authority of the House of Commons. London. The Stationary Office Limited. Available to download at IEA, (2012). [online] Available at: [Accessed 14 Jan. 2013]. Lahr, M. and Stevens, B. (2002). A Study of the Role of Regionalization in the Generation of Aggregation Error in Regional Input -Output Models. Journal of Regional Science, 42(3), pp Legislation.gov.uk. (2016). Climate Change (Scotland) Act [online] Available at: [Accessed 15 May 2016]. Lenzen, M., Murray, J., Sack, F. and Wiedmann, T. (2007). Shared producer and consumer responsibility theory and practice. Ecological Economics, 61(1), pp Lenzen, M., Pade, L. and Munksgaard, J. (2004). CO2 multipliers in multi-region input-output models. Economic Systems Research, 16(4), pp

127 Leontief, W. (1970). Environmental repercussions and the economic structure: an input-output approach. The review of economics and statistics, 52(3), pp Machado, G., Schaeffer, R. and Worrell, E. (2001). Energy and carbon embodied in the international trade of Brazil: an input--output approach. Ecological economics, 39(3), pp McGregor, P., Swales, J. and Turner, K. (2008). The CO2 trade balance between Scotland and the rest of the UK: Performing a multi-region environmental input output analysis with limited data. Ecological Economics, 66(4), pp McNicoll, I.H. and Blackmore, D. (1993). A pilot study on the construction of a Scottish environmental input-output system. Report to Scottish Enterprise. Department of Economics, University of Strathclyde. Midmore, P., Munday, M. and Roberts, A. (2006). Assessing industry linkages using regional input output tables. Regional Studies, 40(3), pp Miller, R. and Blair, P. (1981). Spatial aggregation in interregional input-output models. Papers of the Regional Science Association, 48(1), pp Miller, R. and Blair, P. (2009). Input-output analysis. Cambridge [England]: Cambridge University Press. Miller, R. and Lahr, M. (2001). A taxonomy of extractions. In: Lahr, M. and Miller, R. (2001). Regional science perspectives in economic analysis. Amsterdam: Elsevier, pp

128 Munksgaard, J. and Pedersen, K. (2001). CO2 accounts for open economies: producer or consumer responsibility?. Energy Policy, 29(4), pp Munoz, P. and Steininger, K. (2010). Austria's CO2 responsibility and the carbon content of its international trade. Ecological economics, 69(10), pp OECD, (2015). OECD Inter-Country Input-Output (ICIO) tables. [online] Available at: Okamoto, N. (2005). Agglomeration, Intraregional and Interregional Linkages in China. In: Okamoto, N. and Ihara, T. (2005). Spatial structure and regional development in China: An Interregional Input-Output Approach. Basingstoke [England]: Palgrave Macmillan, pp Rasmussen, P. (1957). Studies in inter-sectoral relations. Amsterdam: North-Holland. Scottish Government, (2014). Input-Output Tables Industry by Industry Table. [online] Available at: Output/Downloads/IO IxI SEPA. Scottish Environment Protection Agency. [online] Available at: Serrano, M. and Dietzenbacher, E. (2010). Responsibility and trade emission balances: An evaluation of approaches. Ecological economics, 69(11), pp

129 Shui, B. and Harriss, R.C. (2006). The role of CO2 embodiment in US China trade. Energy Policy, 34(1), pp Sonis, M., Hewings, J. and Guo, J. (2000). A New Image of Classical Key Sector Analysis: Minimum Information Decomposition of the Leontief Inverse. Economic Systems Research, 12(3), pp Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R. and de Vries, G. J. (2015). An Illustrated User Guide to the World Input Output Database: the Case of Global Automotive Production. Review of International Economics, 23, pp Turner, K. (2006). Additional precision provided by region-specific data: The identification of fuel-use and pollution-generation coefficients in the Jersey economy. Regional Studies, 40(4), pp Turner, K., Lenzen, M., Wiedmann, T. and Barrett, J. (2007). Examining the global environmental impact of regional consumption activities Part 1: A technical note on combining input--output and ecological footprint analysis. Ecological Economics, 62(1), pp Turner, K., Munday, M., McGregor, P. and Swales, K. (2012). How responsible is a region for its carbon emissions? An empirical general equilibrium analysis. Ecological Economics, 76, pp Turner, K., Munday, M., McIntyre, S. and Jensen, C. (2011a). Incorporating jurisdiction issues into regional carbon accounts under production and consumption accounting principles. Environment and Planning A, 43, pp

130 Turner, K., Yamano, N., Druckman, A., Ha, S., de Fence, J., McIntyre, S. and Munday, M. (2011b). An input-output carbon accounting tool: with carbon footprint estimates for the UK and Scotland. Fraser Economic Commentary, Special Issue No1, pp United Nations, (1992). United Nations Framework Convention on Climate Change. [online] Available at: ng.pdf UNFCCC (2014). Greenhouse Gas Inventory Data - Detailed data by Party. [online] Unfccc.int. Available at: [Accessed 19 Sep. 2014]. Wiebe, K., Bruckner, M., Giljum, S. and Lutz, C. (2012). Calculating energy-related CO2 emissions embodied in international trade using a global input-output model. Economic Systems Research, 24(2), pp Wiedmann, T., Lenzen, M., Turner, K. and Barrett, J. (2007). Examining the global environmental impact of regional consumption activities Part 2: Review of input--output models for the assessment of environmental impacts embodied in trade. Ecological Economics, 61(1), pp Wiedmann, T., Wilting, H., Lenzen, M., Lutter, S. and Viveka, P. (2011). Quo Vadis MRIO? Methodological, data and institutional requirements for multi-region input output analysis. Ecological Economics, 70(11), pp Wiedmann, T. (2009). A review of recent multi-region input--output models used for consumptionbased emission and resource accounting. Ecological Economics, 69(2), pp

131 Wyckoff, A. and Roop, J. (1994). The embodiment of carbon in imports of manufactured products: Implications for international agreements on greenhouse gas emissions. Energy policy, 22(3), pp

132 Appendices Appendix 1.A: The Industry Industry industrial sectors of WIOD database 132

133 Appendix 2.A; Table 2.A.1: The countries included in the OECD Inter-Country Input Output Database 133

134 Appendix 2.B; Table 2.B.1: The Industry Industry industrial sectors of the OECD Inter-Country Input Output Database 134